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/md_var.h>
62 #include <machine/pcb.h>
63 #include <machine/psl.h>
64 #include <machine/segments.h>
65 #include <machine/specialreg.h>
66 #include <x86/ifunc.h>
70 #include <vm/vm_map.h>
72 #include <compat/freebsd32/freebsd32_util.h>
73 #include <amd64/linux32/linux.h>
74 #include <amd64/linux32/linux32_proto.h>
75 #include <compat/linux/linux_emul.h>
76 #include <compat/linux/linux_ipc.h>
77 #include <compat/linux/linux_misc.h>
78 #include <compat/linux/linux_mmap.h>
79 #include <compat/linux/linux_signal.h>
80 #include <compat/linux/linux_util.h>
82 static void bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru);
84 struct l_old_select_argv {
88 l_uintptr_t exceptfds;
93 bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru)
96 lru->ru_utime.tv_sec = ru->ru_utime.tv_sec;
97 lru->ru_utime.tv_usec = ru->ru_utime.tv_usec;
98 lru->ru_stime.tv_sec = ru->ru_stime.tv_sec;
99 lru->ru_stime.tv_usec = ru->ru_stime.tv_usec;
100 lru->ru_maxrss = ru->ru_maxrss;
101 lru->ru_ixrss = ru->ru_ixrss;
102 lru->ru_idrss = ru->ru_idrss;
103 lru->ru_isrss = ru->ru_isrss;
104 lru->ru_minflt = ru->ru_minflt;
105 lru->ru_majflt = ru->ru_majflt;
106 lru->ru_nswap = ru->ru_nswap;
107 lru->ru_inblock = ru->ru_inblock;
108 lru->ru_oublock = ru->ru_oublock;
109 lru->ru_msgsnd = ru->ru_msgsnd;
110 lru->ru_msgrcv = ru->ru_msgrcv;
111 lru->ru_nsignals = ru->ru_nsignals;
112 lru->ru_nvcsw = ru->ru_nvcsw;
113 lru->ru_nivcsw = ru->ru_nivcsw;
117 linux_copyout_rusage(struct rusage *ru, void *uaddr)
121 bsd_to_linux_rusage(ru, &lru);
123 return (copyout(&lru, uaddr, sizeof(struct l_rusage)));
127 linux_execve(struct thread *td, struct linux_execve_args *args)
129 struct image_args eargs;
133 LCONVPATHEXIST(td, args->path, &path);
135 error = freebsd32_exec_copyin_args(&eargs, path, UIO_SYSSPACE,
136 args->argp, args->envp);
139 error = linux_common_execve(td, &eargs);
143 CTASSERT(sizeof(struct l_iovec32) == 8);
146 linux32_copyinuio(struct l_iovec32 *iovp, l_ulong iovcnt, struct uio **uiop)
148 struct l_iovec32 iov32;
155 if (iovcnt > UIO_MAXIOV)
157 iovlen = iovcnt * sizeof(struct iovec);
158 uio = malloc(iovlen + sizeof(*uio), M_IOV, M_WAITOK);
159 iov = (struct iovec *)(uio + 1);
160 for (i = 0; i < iovcnt; i++) {
161 error = copyin(&iovp[i], &iov32, sizeof(struct l_iovec32));
166 iov[i].iov_base = PTRIN(iov32.iov_base);
167 iov[i].iov_len = iov32.iov_len;
170 uio->uio_iovcnt = iovcnt;
171 uio->uio_segflg = UIO_USERSPACE;
172 uio->uio_offset = -1;
174 for (i = 0; i < iovcnt; i++) {
175 if (iov->iov_len > INT_MAX - uio->uio_resid) {
179 uio->uio_resid += iov->iov_len;
187 linux32_copyiniov(struct l_iovec32 *iovp32, l_ulong iovcnt, struct iovec **iovp,
190 struct l_iovec32 iov32;
196 if (iovcnt > UIO_MAXIOV)
198 iovlen = iovcnt * sizeof(struct iovec);
199 iov = malloc(iovlen, M_IOV, M_WAITOK);
200 for (i = 0; i < iovcnt; i++) {
201 error = copyin(&iovp32[i], &iov32, sizeof(struct l_iovec32));
206 iov[i].iov_base = PTRIN(iov32.iov_base);
207 iov[i].iov_len = iov32.iov_len;
215 linux_readv(struct thread *td, struct linux_readv_args *uap)
220 error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
223 error = kern_readv(td, uap->fd, auio);
229 linux_writev(struct thread *td, struct linux_writev_args *uap)
234 error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
237 error = kern_writev(td, uap->fd, auio);
242 struct l_ipc_kludge {
248 linux_ipc(struct thread *td, struct linux_ipc_args *args)
251 switch (args->what & 0xFFFF) {
253 struct linux_semop_args a;
255 a.semid = args->arg1;
256 a.tsops = PTRIN(args->ptr);
257 a.nsops = args->arg2;
258 return (linux_semop(td, &a));
261 struct linux_semget_args a;
264 a.nsems = args->arg2;
265 a.semflg = args->arg3;
266 return (linux_semget(td, &a));
269 struct linux_semctl_args a;
272 a.semid = args->arg1;
273 a.semnum = args->arg2;
275 error = copyin(PTRIN(args->ptr), &a.arg, sizeof(a.arg));
278 return (linux_semctl(td, &a));
281 struct linux_msgsnd_args a;
283 a.msqid = args->arg1;
284 a.msgp = PTRIN(args->ptr);
285 a.msgsz = args->arg2;
286 a.msgflg = args->arg3;
287 return (linux_msgsnd(td, &a));
290 struct linux_msgrcv_args a;
292 a.msqid = args->arg1;
293 a.msgsz = args->arg2;
294 a.msgflg = args->arg3;
295 if ((args->what >> 16) == 0) {
296 struct l_ipc_kludge tmp;
301 error = copyin(PTRIN(args->ptr), &tmp, sizeof(tmp));
304 a.msgp = PTRIN(tmp.msgp);
305 a.msgtyp = tmp.msgtyp;
307 a.msgp = PTRIN(args->ptr);
308 a.msgtyp = args->arg5;
310 return (linux_msgrcv(td, &a));
313 struct linux_msgget_args a;
316 a.msgflg = args->arg2;
317 return (linux_msgget(td, &a));
320 struct linux_msgctl_args a;
322 a.msqid = args->arg1;
324 a.buf = PTRIN(args->ptr);
325 return (linux_msgctl(td, &a));
328 struct linux_shmat_args a;
332 a.shmid = args->arg1;
333 a.shmaddr = PTRIN(args->ptr);
334 a.shmflg = args->arg2;
335 error = linux_shmat(td, &a);
338 addr = td->td_retval[0];
339 error = copyout(&addr, PTRIN(args->arg3), sizeof(addr));
340 td->td_retval[0] = 0;
344 struct linux_shmdt_args a;
346 a.shmaddr = PTRIN(args->ptr);
347 return (linux_shmdt(td, &a));
350 struct linux_shmget_args a;
354 a.shmflg = args->arg3;
355 return (linux_shmget(td, &a));
358 struct linux_shmctl_args a;
360 a.shmid = args->arg1;
362 a.buf = PTRIN(args->ptr);
363 return (linux_shmctl(td, &a));
373 linux_old_select(struct thread *td, struct linux_old_select_args *args)
375 struct l_old_select_argv linux_args;
376 struct linux_select_args newsel;
379 error = copyin(args->ptr, &linux_args, sizeof(linux_args));
383 newsel.nfds = linux_args.nfds;
384 newsel.readfds = PTRIN(linux_args.readfds);
385 newsel.writefds = PTRIN(linux_args.writefds);
386 newsel.exceptfds = PTRIN(linux_args.exceptfds);
387 newsel.timeout = PTRIN(linux_args.timeout);
388 return (linux_select(td, &newsel));
392 linux_set_cloned_tls(struct thread *td, void *desc)
394 struct user_segment_descriptor sd;
395 struct l_user_desc info;
400 error = copyin(desc, &info, sizeof(struct l_user_desc));
402 linux_msg(td, "set_cloned_tls copyin info failed!");
404 /* We might copy out the entry_number as GUGS32_SEL. */
405 info.entry_number = GUGS32_SEL;
406 error = copyout(&info, desc, sizeof(struct l_user_desc));
408 linux_msg(td, "set_cloned_tls copyout info failed!");
410 a[0] = LINUX_LDT_entry_a(&info);
411 a[1] = LINUX_LDT_entry_b(&info);
413 memcpy(&sd, &a, sizeof(a));
415 pcb->pcb_gsbase = (register_t)info.base_addr;
416 td->td_frame->tf_gs = GSEL(GUGS32_SEL, SEL_UPL);
417 set_pcb_flags(pcb, PCB_32BIT);
424 linux_set_upcall_kse(struct thread *td, register_t stack)
428 td->td_frame->tf_rsp = stack;
431 * The newly created Linux thread returns
432 * to the user space by the same path that a parent do.
434 td->td_frame->tf_rax = 0;
439 linux_mmap2(struct thread *td, struct linux_mmap2_args *args)
442 return (linux_mmap_common(td, PTROUT(args->addr), args->len, args->prot,
443 args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff *
448 linux_mmap(struct thread *td, struct linux_mmap_args *args)
451 struct l_mmap_argv linux_args;
453 error = copyin(args->ptr, &linux_args, sizeof(linux_args));
457 return (linux_mmap_common(td, linux_args.addr, linux_args.len,
458 linux_args.prot, linux_args.flags, linux_args.fd,
459 (uint32_t)linux_args.pgoff));
463 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
466 return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len, uap->prot));
470 linux_madvise(struct thread *td, struct linux_madvise_args *uap)
473 return (linux_madvise_common(td, PTROUT(uap->addr), uap->len, uap->behav));
477 linux_iopl(struct thread *td, struct linux_iopl_args *args)
481 if (args->level < 0 || args->level > 3)
483 if ((error = priv_check(td, PRIV_IO)) != 0)
485 if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
487 td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) |
488 (args->level * (PSL_IOPL / 3));
494 linux_sigaction(struct thread *td, struct linux_sigaction_args *args)
497 l_sigaction_t act, oact;
500 if (args->nsa != NULL) {
501 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
504 act.lsa_handler = osa.lsa_handler;
505 act.lsa_flags = osa.lsa_flags;
506 act.lsa_restorer = osa.lsa_restorer;
507 LINUX_SIGEMPTYSET(act.lsa_mask);
508 act.lsa_mask.__mask = osa.lsa_mask;
511 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL,
512 args->osa ? &oact : NULL);
514 if (args->osa != NULL && !error) {
515 osa.lsa_handler = oact.lsa_handler;
516 osa.lsa_flags = oact.lsa_flags;
517 osa.lsa_restorer = oact.lsa_restorer;
518 osa.lsa_mask = oact.lsa_mask.__mask;
519 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
526 * Linux has two extra args, restart and oldmask. We don't use these,
527 * but it seems that "restart" is actually a context pointer that
528 * enables the signal to happen with a different register set.
531 linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args)
536 LINUX_SIGEMPTYSET(mask);
537 mask.__mask = args->mask;
538 linux_to_bsd_sigset(&mask, &sigmask);
539 return (kern_sigsuspend(td, sigmask));
543 linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap)
549 if (uap->sigsetsize != sizeof(l_sigset_t))
552 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t));
556 linux_to_bsd_sigset(&lmask, &sigmask);
557 return (kern_sigsuspend(td, sigmask));
561 linux_pause(struct thread *td, struct linux_pause_args *args)
563 struct proc *p = td->td_proc;
567 sigmask = td->td_sigmask;
569 return (kern_sigsuspend(td, sigmask));
573 linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap)
579 if (uap->uss != NULL) {
580 error = copyin(uap->uss, &lss, sizeof(l_stack_t));
584 ss.ss_sp = PTRIN(lss.ss_sp);
585 ss.ss_size = lss.ss_size;
586 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags);
588 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL,
589 (uap->uoss != NULL) ? &oss : NULL);
590 if (!error && uap->uoss != NULL) {
591 lss.ss_sp = PTROUT(oss.ss_sp);
592 lss.ss_size = oss.ss_size;
593 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
594 error = copyout(&lss, uap->uoss, sizeof(l_stack_t));
601 linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap)
610 atv32.tv_sec = atv.tv_sec;
611 atv32.tv_usec = atv.tv_usec;
612 error = copyout(&atv32, uap->tp, sizeof(atv32));
614 if (error == 0 && uap->tzp != NULL) {
615 rtz.tz_minuteswest = 0;
617 error = copyout(&rtz, uap->tzp, sizeof(rtz));
623 linux_settimeofday(struct thread *td, struct linux_settimeofday_args *uap)
626 struct timeval atv, *tvp;
627 struct timezone atz, *tzp;
631 error = copyin(uap->tp, &atv32, sizeof(atv32));
634 atv.tv_sec = atv32.tv_sec;
635 atv.tv_usec = atv32.tv_usec;
640 error = copyin(uap->tzp, &atz, sizeof(atz));
646 return (kern_settimeofday(td, tvp, tzp));
650 linux_getrusage(struct thread *td, struct linux_getrusage_args *uap)
655 error = kern_getrusage(td, uap->who, &s);
658 if (uap->rusage != NULL)
659 error = linux_copyout_rusage(&s, uap->rusage);
664 linux_set_thread_area(struct thread *td,
665 struct linux_set_thread_area_args *args)
667 struct l_user_desc info;
668 struct user_segment_descriptor sd;
673 error = copyin(args->desc, &info, sizeof(struct l_user_desc));
678 * Semantics of Linux version: every thread in the system has array
679 * of three TLS descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown.
680 * This syscall loads one of the selected TLS decriptors with a value
681 * and also loads GDT descriptors 6, 7 and 8 with the content of
682 * the per-thread descriptors.
684 * Semantics of FreeBSD version: I think we can ignore that Linux has
685 * three per-thread descriptors and use just the first one.
686 * The tls_array[] is used only in [gs]et_thread_area() syscalls and
687 * for loading the GDT descriptors. We use just one GDT descriptor
688 * for TLS, so we will load just one.
690 * XXX: This doesn't work when a user space process tries to use more
691 * than one TLS segment. Comment in the Linux source says wine might
696 * GLIBC reads current %gs and call set_thread_area() with it.
697 * We should let GUDATA_SEL and GUGS32_SEL proceed as well because
698 * we use these segments.
700 switch (info.entry_number) {
705 info.entry_number = GUGS32_SEL;
712 * We have to copy out the GDT entry we use.
714 * XXX: What if a user space program does not check the return value
715 * and tries to use 6, 7 or 8?
717 error = copyout(&info, args->desc, sizeof(struct l_user_desc));
721 if (LINUX_LDT_empty(&info)) {
725 a[0] = LINUX_LDT_entry_a(&info);
726 a[1] = LINUX_LDT_entry_b(&info);
729 memcpy(&sd, &a, sizeof(a));
731 pcb->pcb_gsbase = (register_t)info.base_addr;
732 set_pcb_flags(pcb, PCB_32BIT);
733 update_gdt_gsbase(td, info.base_addr);
738 int futex_xchgl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
739 int futex_xchgl_smap(int oparg, uint32_t *uaddr, int *oldval);
740 DEFINE_IFUNC(, int, futex_xchgl, (int, uint32_t *, int *))
743 return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
744 futex_xchgl_smap : futex_xchgl_nosmap);
747 int futex_addl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
748 int futex_addl_smap(int oparg, uint32_t *uaddr, int *oldval);
749 DEFINE_IFUNC(, int, futex_addl, (int, uint32_t *, int *))
752 return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
753 futex_addl_smap : futex_addl_nosmap);
756 int futex_orl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
757 int futex_orl_smap(int oparg, uint32_t *uaddr, int *oldval);
758 DEFINE_IFUNC(, int, futex_orl, (int, uint32_t *, int *))
761 return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
762 futex_orl_smap : futex_orl_nosmap);
765 int futex_andl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
766 int futex_andl_smap(int oparg, uint32_t *uaddr, int *oldval);
767 DEFINE_IFUNC(, int, futex_andl, (int, uint32_t *, int *))
770 return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
771 futex_andl_smap : futex_andl_nosmap);
774 int futex_xorl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
775 int futex_xorl_smap(int oparg, uint32_t *uaddr, int *oldval);
776 DEFINE_IFUNC(, int, futex_xorl, (int, uint32_t *, int *))
779 return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
780 futex_xorl_smap : futex_xorl_nosmap);