2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 2004 Tim J. Robbins
5 * Copyright (c) 2002 Doug Rabson
6 * Copyright (c) 2000 Marcel Moolenaar
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer
14 * in this position and unchanged.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. The name of the author may not be used to endorse or promote products
19 * derived from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
22 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
23 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
24 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
26 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
27 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
28 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
29 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
30 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
36 #include "opt_compat.h"
38 #include <sys/param.h>
39 #include <sys/capsicum.h>
40 #include <sys/clock.h>
41 #include <sys/fcntl.h>
43 #include <sys/imgact.h>
44 #include <sys/kernel.h>
45 #include <sys/limits.h>
47 #include <sys/malloc.h>
49 #include <sys/mutex.h>
52 #include <sys/resource.h>
53 #include <sys/resourcevar.h>
54 #include <sys/syscallsubr.h>
55 #include <sys/sysproto.h>
56 #include <sys/systm.h>
57 #include <sys/unistd.h>
60 #include <machine/frame.h>
61 #include <machine/pcb.h>
62 #include <machine/psl.h>
63 #include <machine/segments.h>
64 #include <machine/specialreg.h>
68 #include <vm/vm_map.h>
70 #include <compat/freebsd32/freebsd32_util.h>
71 #include <amd64/linux32/linux.h>
72 #include <amd64/linux32/linux32_proto.h>
73 #include <compat/linux/linux_emul.h>
74 #include <compat/linux/linux_ipc.h>
75 #include <compat/linux/linux_misc.h>
76 #include <compat/linux/linux_mmap.h>
77 #include <compat/linux/linux_signal.h>
78 #include <compat/linux/linux_util.h>
80 static void bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru);
82 struct l_old_select_argv {
86 l_uintptr_t exceptfds;
92 bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru)
95 lru->ru_utime.tv_sec = ru->ru_utime.tv_sec;
96 lru->ru_utime.tv_usec = ru->ru_utime.tv_usec;
97 lru->ru_stime.tv_sec = ru->ru_stime.tv_sec;
98 lru->ru_stime.tv_usec = ru->ru_stime.tv_usec;
99 lru->ru_maxrss = ru->ru_maxrss;
100 lru->ru_ixrss = ru->ru_ixrss;
101 lru->ru_idrss = ru->ru_idrss;
102 lru->ru_isrss = ru->ru_isrss;
103 lru->ru_minflt = ru->ru_minflt;
104 lru->ru_majflt = ru->ru_majflt;
105 lru->ru_nswap = ru->ru_nswap;
106 lru->ru_inblock = ru->ru_inblock;
107 lru->ru_oublock = ru->ru_oublock;
108 lru->ru_msgsnd = ru->ru_msgsnd;
109 lru->ru_msgrcv = ru->ru_msgrcv;
110 lru->ru_nsignals = ru->ru_nsignals;
111 lru->ru_nvcsw = ru->ru_nvcsw;
112 lru->ru_nivcsw = ru->ru_nivcsw;
116 linux_copyout_rusage(struct rusage *ru, void *uaddr)
120 bsd_to_linux_rusage(ru, &lru);
122 return (copyout(&lru, uaddr, sizeof(struct l_rusage)));
126 linux_execve(struct thread *td, struct linux_execve_args *args)
128 struct image_args eargs;
132 LCONVPATHEXIST(td, args->path, &path);
136 printf(ARGS(execve, "%s"), path);
139 error = freebsd32_exec_copyin_args(&eargs, path, UIO_SYSSPACE,
140 args->argp, args->envp);
143 error = linux_common_execve(td, &eargs);
147 CTASSERT(sizeof(struct l_iovec32) == 8);
150 linux32_copyinuio(struct l_iovec32 *iovp, l_ulong iovcnt, struct uio **uiop)
152 struct l_iovec32 iov32;
159 if (iovcnt > UIO_MAXIOV)
161 iovlen = iovcnt * sizeof(struct iovec);
162 uio = malloc(iovlen + sizeof(*uio), M_IOV, M_WAITOK);
163 iov = (struct iovec *)(uio + 1);
164 for (i = 0; i < iovcnt; i++) {
165 error = copyin(&iovp[i], &iov32, sizeof(struct l_iovec32));
170 iov[i].iov_base = PTRIN(iov32.iov_base);
171 iov[i].iov_len = iov32.iov_len;
174 uio->uio_iovcnt = iovcnt;
175 uio->uio_segflg = UIO_USERSPACE;
176 uio->uio_offset = -1;
178 for (i = 0; i < iovcnt; i++) {
179 if (iov->iov_len > INT_MAX - uio->uio_resid) {
183 uio->uio_resid += iov->iov_len;
191 linux32_copyiniov(struct l_iovec32 *iovp32, l_ulong iovcnt, struct iovec **iovp,
194 struct l_iovec32 iov32;
200 if (iovcnt > UIO_MAXIOV)
202 iovlen = iovcnt * sizeof(struct iovec);
203 iov = malloc(iovlen, M_IOV, M_WAITOK);
204 for (i = 0; i < iovcnt; i++) {
205 error = copyin(&iovp32[i], &iov32, sizeof(struct l_iovec32));
210 iov[i].iov_base = PTRIN(iov32.iov_base);
211 iov[i].iov_len = iov32.iov_len;
219 linux_readv(struct thread *td, struct linux_readv_args *uap)
224 error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
227 error = kern_readv(td, uap->fd, auio);
233 linux_writev(struct thread *td, struct linux_writev_args *uap)
238 error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
241 error = kern_writev(td, uap->fd, auio);
246 struct l_ipc_kludge {
252 linux_ipc(struct thread *td, struct linux_ipc_args *args)
255 switch (args->what & 0xFFFF) {
257 struct linux_semop_args a;
259 a.semid = args->arg1;
261 a.nsops = args->arg2;
262 return (linux_semop(td, &a));
265 struct linux_semget_args a;
268 a.nsems = args->arg2;
269 a.semflg = args->arg3;
270 return (linux_semget(td, &a));
273 struct linux_semctl_args a;
276 a.semid = args->arg1;
277 a.semnum = args->arg2;
279 error = copyin(args->ptr, &a.arg, sizeof(a.arg));
282 return (linux_semctl(td, &a));
285 struct linux_msgsnd_args a;
287 a.msqid = args->arg1;
289 a.msgsz = args->arg2;
290 a.msgflg = args->arg3;
291 return (linux_msgsnd(td, &a));
294 struct linux_msgrcv_args a;
296 a.msqid = args->arg1;
297 a.msgsz = args->arg2;
298 a.msgflg = args->arg3;
299 if ((args->what >> 16) == 0) {
300 struct l_ipc_kludge tmp;
305 error = copyin(args->ptr, &tmp, sizeof(tmp));
308 a.msgp = PTRIN(tmp.msgp);
309 a.msgtyp = tmp.msgtyp;
312 a.msgtyp = args->arg5;
314 return (linux_msgrcv(td, &a));
317 struct linux_msgget_args a;
320 a.msgflg = args->arg2;
321 return (linux_msgget(td, &a));
324 struct linux_msgctl_args a;
326 a.msqid = args->arg1;
329 return (linux_msgctl(td, &a));
332 struct linux_shmat_args a;
334 a.shmid = args->arg1;
335 a.shmaddr = args->ptr;
336 a.shmflg = args->arg2;
337 a.raddr = PTRIN((l_uint)args->arg3);
338 return (linux_shmat(td, &a));
341 struct linux_shmdt_args a;
343 a.shmaddr = args->ptr;
344 return (linux_shmdt(td, &a));
347 struct linux_shmget_args a;
351 a.shmflg = args->arg3;
352 return (linux_shmget(td, &a));
355 struct linux_shmctl_args a;
357 a.shmid = args->arg1;
360 return (linux_shmctl(td, &a));
370 linux_old_select(struct thread *td, struct linux_old_select_args *args)
372 struct l_old_select_argv linux_args;
373 struct linux_select_args newsel;
377 if (ldebug(old_select))
378 printf(ARGS(old_select, "%p"), args->ptr);
381 error = copyin(args->ptr, &linux_args, sizeof(linux_args));
385 newsel.nfds = linux_args.nfds;
386 newsel.readfds = PTRIN(linux_args.readfds);
387 newsel.writefds = PTRIN(linux_args.writefds);
388 newsel.exceptfds = PTRIN(linux_args.exceptfds);
389 newsel.timeout = PTRIN(linux_args.timeout);
390 return (linux_select(td, &newsel));
394 linux_set_cloned_tls(struct thread *td, void *desc)
396 struct user_segment_descriptor sd;
397 struct l_user_desc info;
402 error = copyin(desc, &info, sizeof(struct l_user_desc));
404 printf(LMSG("copyin failed!"));
406 /* We might copy out the entry_number as GUGS32_SEL. */
407 info.entry_number = GUGS32_SEL;
408 error = copyout(&info, desc, sizeof(struct l_user_desc));
410 printf(LMSG("copyout failed!"));
412 a[0] = LINUX_LDT_entry_a(&info);
413 a[1] = LINUX_LDT_entry_b(&info);
415 memcpy(&sd, &a, sizeof(a));
418 printf("Segment created in clone with "
419 "CLONE_SETTLS: lobase: %x, hibase: %x, "
420 "lolimit: %x, hilimit: %x, type: %i, "
421 "dpl: %i, p: %i, xx: %i, long: %i, "
422 "def32: %i, gran: %i\n", sd.sd_lobase,
423 sd.sd_hibase, sd.sd_lolimit, sd.sd_hilimit,
424 sd.sd_type, sd.sd_dpl, sd.sd_p, sd.sd_xx,
425 sd.sd_long, sd.sd_def32, sd.sd_gran);
428 pcb->pcb_gsbase = (register_t)info.base_addr;
429 td->td_frame->tf_gs = GSEL(GUGS32_SEL, SEL_UPL);
430 set_pcb_flags(pcb, PCB_32BIT);
437 linux_set_upcall_kse(struct thread *td, register_t stack)
441 td->td_frame->tf_rsp = stack;
444 * The newly created Linux thread returns
445 * to the user space by the same path that a parent do.
447 td->td_frame->tf_rax = 0;
452 linux_mmap2(struct thread *td, struct linux_mmap2_args *args)
457 printf(ARGS(mmap2, "0x%08x, %d, %d, 0x%08x, %d, %d"),
458 args->addr, args->len, args->prot,
459 args->flags, args->fd, args->pgoff);
462 return (linux_mmap_common(td, PTROUT(args->addr), args->len, args->prot,
463 args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff *
468 linux_mmap(struct thread *td, struct linux_mmap_args *args)
471 struct l_mmap_argv linux_args;
473 error = copyin(args->ptr, &linux_args, sizeof(linux_args));
479 printf(ARGS(mmap, "0x%08x, %d, %d, 0x%08x, %d, %d"),
480 linux_args.addr, linux_args.len, linux_args.prot,
481 linux_args.flags, linux_args.fd, linux_args.pgoff);
484 return (linux_mmap_common(td, linux_args.addr, linux_args.len,
485 linux_args.prot, linux_args.flags, linux_args.fd,
486 (uint32_t)linux_args.pgoff));
490 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
493 return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len, uap->prot));
497 linux_iopl(struct thread *td, struct linux_iopl_args *args)
501 if (args->level < 0 || args->level > 3)
503 if ((error = priv_check(td, PRIV_IO)) != 0)
505 if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
507 td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) |
508 (args->level * (PSL_IOPL / 3));
514 linux_sigaction(struct thread *td, struct linux_sigaction_args *args)
517 l_sigaction_t act, oact;
521 if (ldebug(sigaction))
522 printf(ARGS(sigaction, "%d, %p, %p"),
523 args->sig, (void *)args->nsa, (void *)args->osa);
526 if (args->nsa != NULL) {
527 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
530 act.lsa_handler = osa.lsa_handler;
531 act.lsa_flags = osa.lsa_flags;
532 act.lsa_restorer = osa.lsa_restorer;
533 LINUX_SIGEMPTYSET(act.lsa_mask);
534 act.lsa_mask.__mask = osa.lsa_mask;
537 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL,
538 args->osa ? &oact : NULL);
540 if (args->osa != NULL && !error) {
541 osa.lsa_handler = oact.lsa_handler;
542 osa.lsa_flags = oact.lsa_flags;
543 osa.lsa_restorer = oact.lsa_restorer;
544 osa.lsa_mask = oact.lsa_mask.__mask;
545 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
552 * Linux has two extra args, restart and oldmask. We don't use these,
553 * but it seems that "restart" is actually a context pointer that
554 * enables the signal to happen with a different register set.
557 linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args)
563 if (ldebug(sigsuspend))
564 printf(ARGS(sigsuspend, "%08lx"), (unsigned long)args->mask);
567 LINUX_SIGEMPTYSET(mask);
568 mask.__mask = args->mask;
569 linux_to_bsd_sigset(&mask, &sigmask);
570 return (kern_sigsuspend(td, sigmask));
574 linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap)
581 if (ldebug(rt_sigsuspend))
582 printf(ARGS(rt_sigsuspend, "%p, %d"),
583 (void *)uap->newset, uap->sigsetsize);
586 if (uap->sigsetsize != sizeof(l_sigset_t))
589 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t));
593 linux_to_bsd_sigset(&lmask, &sigmask);
594 return (kern_sigsuspend(td, sigmask));
598 linux_pause(struct thread *td, struct linux_pause_args *args)
600 struct proc *p = td->td_proc;
605 printf(ARGS(pause, ""));
609 sigmask = td->td_sigmask;
611 return (kern_sigsuspend(td, sigmask));
615 linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap)
622 if (ldebug(sigaltstack))
623 printf(ARGS(sigaltstack, "%p, %p"), uap->uss, uap->uoss);
626 if (uap->uss != NULL) {
627 error = copyin(uap->uss, &lss, sizeof(l_stack_t));
631 ss.ss_sp = PTRIN(lss.ss_sp);
632 ss.ss_size = lss.ss_size;
633 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags);
635 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL,
636 (uap->uoss != NULL) ? &oss : NULL);
637 if (!error && uap->uoss != NULL) {
638 lss.ss_sp = PTROUT(oss.ss_sp);
639 lss.ss_size = oss.ss_size;
640 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
641 error = copyout(&lss, uap->uoss, sizeof(l_stack_t));
648 linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args)
652 if (ldebug(ftruncate64))
653 printf(ARGS(ftruncate64, "%u, %jd"), args->fd,
654 (intmax_t)args->length);
657 return (kern_ftruncate(td, args->fd, args->length));
661 linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap)
670 atv32.tv_sec = atv.tv_sec;
671 atv32.tv_usec = atv.tv_usec;
672 error = copyout(&atv32, uap->tp, sizeof(atv32));
674 if (error == 0 && uap->tzp != NULL) {
675 rtz.tz_minuteswest = tz_minuteswest;
676 rtz.tz_dsttime = tz_dsttime;
677 error = copyout(&rtz, uap->tzp, sizeof(rtz));
683 linux_settimeofday(struct thread *td, struct linux_settimeofday_args *uap)
686 struct timeval atv, *tvp;
687 struct timezone atz, *tzp;
691 error = copyin(uap->tp, &atv32, sizeof(atv32));
694 atv.tv_sec = atv32.tv_sec;
695 atv.tv_usec = atv32.tv_usec;
700 error = copyin(uap->tzp, &atz, sizeof(atz));
706 return (kern_settimeofday(td, tvp, tzp));
710 linux_getrusage(struct thread *td, struct linux_getrusage_args *uap)
715 error = kern_getrusage(td, uap->who, &s);
718 if (uap->rusage != NULL)
719 error = linux_copyout_rusage(&s, uap->rusage);
724 linux_set_thread_area(struct thread *td,
725 struct linux_set_thread_area_args *args)
727 struct l_user_desc info;
728 struct user_segment_descriptor sd;
733 error = copyin(args->desc, &info, sizeof(struct l_user_desc));
738 if (ldebug(set_thread_area))
739 printf(ARGS(set_thread_area, "%i, %x, %x, %i, %i, %i, "
740 "%i, %i, %i"), info.entry_number, info.base_addr,
741 info.limit, info.seg_32bit, info.contents,
742 info.read_exec_only, info.limit_in_pages,
743 info.seg_not_present, info.useable);
747 * Semantics of Linux version: every thread in the system has array
748 * of three TLS descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown.
749 * This syscall loads one of the selected TLS decriptors with a value
750 * and also loads GDT descriptors 6, 7 and 8 with the content of
751 * the per-thread descriptors.
753 * Semantics of FreeBSD version: I think we can ignore that Linux has
754 * three per-thread descriptors and use just the first one.
755 * The tls_array[] is used only in [gs]et_thread_area() syscalls and
756 * for loading the GDT descriptors. We use just one GDT descriptor
757 * for TLS, so we will load just one.
759 * XXX: This doesn't work when a user space process tries to use more
760 * than one TLS segment. Comment in the Linux source says wine might
765 * GLIBC reads current %gs and call set_thread_area() with it.
766 * We should let GUDATA_SEL and GUGS32_SEL proceed as well because
767 * we use these segments.
769 switch (info.entry_number) {
774 info.entry_number = GUGS32_SEL;
781 * We have to copy out the GDT entry we use.
783 * XXX: What if a user space program does not check the return value
784 * and tries to use 6, 7 or 8?
786 error = copyout(&info, args->desc, sizeof(struct l_user_desc));
790 if (LINUX_LDT_empty(&info)) {
794 a[0] = LINUX_LDT_entry_a(&info);
795 a[1] = LINUX_LDT_entry_b(&info);
798 memcpy(&sd, &a, sizeof(a));
800 if (ldebug(set_thread_area))
801 printf("Segment created in set_thread_area: "
802 "lobase: %x, hibase: %x, lolimit: %x, hilimit: %x, "
803 "type: %i, dpl: %i, p: %i, xx: %i, long: %i, "
804 "def32: %i, gran: %i\n",
819 pcb->pcb_gsbase = (register_t)info.base_addr;
820 set_pcb_flags(pcb, PCB_32BIT);
821 update_gdt_gsbase(td, info.base_addr);