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 <security/audit/audit.h>
74 #include <compat/freebsd32/freebsd32_util.h>
75 #include <amd64/linux32/linux.h>
76 #include <amd64/linux32/linux32_proto.h>
77 #include <compat/linux/linux_emul.h>
78 #include <compat/linux/linux_ipc.h>
79 #include <compat/linux/linux_misc.h>
80 #include <compat/linux/linux_mmap.h>
81 #include <compat/linux/linux_signal.h>
82 #include <compat/linux/linux_util.h>
84 static void bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru);
86 struct l_old_select_argv {
90 l_uintptr_t exceptfds;
95 bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru)
98 lru->ru_utime.tv_sec = ru->ru_utime.tv_sec;
99 lru->ru_utime.tv_usec = ru->ru_utime.tv_usec;
100 lru->ru_stime.tv_sec = ru->ru_stime.tv_sec;
101 lru->ru_stime.tv_usec = ru->ru_stime.tv_usec;
102 lru->ru_maxrss = ru->ru_maxrss;
103 lru->ru_ixrss = ru->ru_ixrss;
104 lru->ru_idrss = ru->ru_idrss;
105 lru->ru_isrss = ru->ru_isrss;
106 lru->ru_minflt = ru->ru_minflt;
107 lru->ru_majflt = ru->ru_majflt;
108 lru->ru_nswap = ru->ru_nswap;
109 lru->ru_inblock = ru->ru_inblock;
110 lru->ru_oublock = ru->ru_oublock;
111 lru->ru_msgsnd = ru->ru_msgsnd;
112 lru->ru_msgrcv = ru->ru_msgrcv;
113 lru->ru_nsignals = ru->ru_nsignals;
114 lru->ru_nvcsw = ru->ru_nvcsw;
115 lru->ru_nivcsw = ru->ru_nivcsw;
119 linux_copyout_rusage(struct rusage *ru, void *uaddr)
123 bsd_to_linux_rusage(ru, &lru);
125 return (copyout(&lru, uaddr, sizeof(struct l_rusage)));
129 linux_execve(struct thread *td, struct linux_execve_args *args)
131 struct image_args eargs;
135 LCONVPATHEXIST(td, args->path, &path);
137 error = freebsd32_exec_copyin_args(&eargs, path, UIO_SYSSPACE,
138 args->argp, args->envp);
141 error = linux_common_execve(td, &eargs);
142 AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td);
146 CTASSERT(sizeof(struct l_iovec32) == 8);
149 linux32_copyinuio(struct l_iovec32 *iovp, l_ulong iovcnt, struct uio **uiop)
151 struct l_iovec32 iov32;
158 if (iovcnt > UIO_MAXIOV)
160 iovlen = iovcnt * sizeof(struct iovec);
161 uio = malloc(iovlen + sizeof(*uio), M_IOV, M_WAITOK);
162 iov = (struct iovec *)(uio + 1);
163 for (i = 0; i < iovcnt; i++) {
164 error = copyin(&iovp[i], &iov32, sizeof(struct l_iovec32));
169 iov[i].iov_base = PTRIN(iov32.iov_base);
170 iov[i].iov_len = iov32.iov_len;
173 uio->uio_iovcnt = iovcnt;
174 uio->uio_segflg = UIO_USERSPACE;
175 uio->uio_offset = -1;
177 for (i = 0; i < iovcnt; i++) {
178 if (iov->iov_len > INT_MAX - uio->uio_resid) {
182 uio->uio_resid += iov->iov_len;
190 linux32_copyiniov(struct l_iovec32 *iovp32, l_ulong iovcnt, struct iovec **iovp,
193 struct l_iovec32 iov32;
199 if (iovcnt > UIO_MAXIOV)
201 iovlen = iovcnt * sizeof(struct iovec);
202 iov = malloc(iovlen, M_IOV, M_WAITOK);
203 for (i = 0; i < iovcnt; i++) {
204 error = copyin(&iovp32[i], &iov32, sizeof(struct l_iovec32));
209 iov[i].iov_base = PTRIN(iov32.iov_base);
210 iov[i].iov_len = iov32.iov_len;
218 linux_readv(struct thread *td, struct linux_readv_args *uap)
223 error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
226 error = kern_readv(td, uap->fd, auio);
232 linux_writev(struct thread *td, struct linux_writev_args *uap)
237 error = linux32_copyinuio(uap->iovp, uap->iovcnt, &auio);
240 error = kern_writev(td, uap->fd, auio);
245 struct l_ipc_kludge {
251 linux_ipc(struct thread *td, struct linux_ipc_args *args)
254 switch (args->what & 0xFFFF) {
256 struct linux_semop_args a;
258 a.semid = args->arg1;
259 a.tsops = PTRIN(args->ptr);
260 a.nsops = args->arg2;
261 return (linux_semop(td, &a));
264 struct linux_semget_args a;
267 a.nsems = args->arg2;
268 a.semflg = args->arg3;
269 return (linux_semget(td, &a));
272 struct linux_semctl_args a;
275 a.semid = args->arg1;
276 a.semnum = args->arg2;
278 error = copyin(PTRIN(args->ptr), &a.arg, sizeof(a.arg));
281 return (linux_semctl(td, &a));
284 struct linux_msgsnd_args a;
286 a.msqid = args->arg1;
287 a.msgp = PTRIN(args->ptr);
288 a.msgsz = args->arg2;
289 a.msgflg = args->arg3;
290 return (linux_msgsnd(td, &a));
293 struct linux_msgrcv_args a;
295 a.msqid = args->arg1;
296 a.msgsz = args->arg2;
297 a.msgflg = args->arg3;
298 if ((args->what >> 16) == 0) {
299 struct l_ipc_kludge tmp;
304 error = copyin(PTRIN(args->ptr), &tmp, sizeof(tmp));
307 a.msgp = PTRIN(tmp.msgp);
308 a.msgtyp = tmp.msgtyp;
310 a.msgp = PTRIN(args->ptr);
311 a.msgtyp = args->arg5;
313 return (linux_msgrcv(td, &a));
316 struct linux_msgget_args a;
319 a.msgflg = args->arg2;
320 return (linux_msgget(td, &a));
323 struct linux_msgctl_args a;
325 a.msqid = args->arg1;
327 a.buf = PTRIN(args->ptr);
328 return (linux_msgctl(td, &a));
331 struct linux_shmat_args a;
335 a.shmid = args->arg1;
336 a.shmaddr = PTRIN(args->ptr);
337 a.shmflg = args->arg2;
338 error = linux_shmat(td, &a);
341 addr = td->td_retval[0];
342 error = copyout(&addr, PTRIN(args->arg3), sizeof(addr));
343 td->td_retval[0] = 0;
347 struct linux_shmdt_args a;
349 a.shmaddr = PTRIN(args->ptr);
350 return (linux_shmdt(td, &a));
353 struct linux_shmget_args a;
357 a.shmflg = args->arg3;
358 return (linux_shmget(td, &a));
361 struct linux_shmctl_args a;
363 a.shmid = args->arg1;
365 a.buf = PTRIN(args->ptr);
366 return (linux_shmctl(td, &a));
376 linux_old_select(struct thread *td, struct linux_old_select_args *args)
378 struct l_old_select_argv linux_args;
379 struct linux_select_args newsel;
382 error = copyin(args->ptr, &linux_args, sizeof(linux_args));
386 newsel.nfds = linux_args.nfds;
387 newsel.readfds = PTRIN(linux_args.readfds);
388 newsel.writefds = PTRIN(linux_args.writefds);
389 newsel.exceptfds = PTRIN(linux_args.exceptfds);
390 newsel.timeout = PTRIN(linux_args.timeout);
391 return (linux_select(td, &newsel));
395 linux_set_cloned_tls(struct thread *td, void *desc)
397 struct user_segment_descriptor sd;
398 struct l_user_desc info;
403 error = copyin(desc, &info, sizeof(struct l_user_desc));
405 linux_msg(td, "set_cloned_tls copyin info failed!");
407 /* We might copy out the entry_number as GUGS32_SEL. */
408 info.entry_number = GUGS32_SEL;
409 error = copyout(&info, desc, sizeof(struct l_user_desc));
411 linux_msg(td, "set_cloned_tls copyout info failed!");
413 a[0] = LINUX_LDT_entry_a(&info);
414 a[1] = LINUX_LDT_entry_b(&info);
416 memcpy(&sd, &a, sizeof(a));
418 pcb->pcb_gsbase = (register_t)info.base_addr;
419 td->td_frame->tf_gs = GSEL(GUGS32_SEL, SEL_UPL);
420 set_pcb_flags(pcb, PCB_32BIT);
427 linux_set_upcall_kse(struct thread *td, register_t stack)
431 td->td_frame->tf_rsp = stack;
434 * The newly created Linux thread returns
435 * to the user space by the same path that a parent do.
437 td->td_frame->tf_rax = 0;
442 linux_mmap2(struct thread *td, struct linux_mmap2_args *args)
445 return (linux_mmap_common(td, PTROUT(args->addr), args->len, args->prot,
446 args->flags, args->fd, (uint64_t)(uint32_t)args->pgoff *
451 linux_mmap(struct thread *td, struct linux_mmap_args *args)
454 struct l_mmap_argv linux_args;
456 error = copyin(args->ptr, &linux_args, sizeof(linux_args));
460 return (linux_mmap_common(td, linux_args.addr, linux_args.len,
461 linux_args.prot, linux_args.flags, linux_args.fd,
462 (uint32_t)linux_args.pgoff));
466 linux_mprotect(struct thread *td, struct linux_mprotect_args *uap)
469 return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len, uap->prot));
473 linux_madvise(struct thread *td, struct linux_madvise_args *uap)
476 return (linux_madvise_common(td, PTROUT(uap->addr), uap->len, uap->behav));
480 linux_iopl(struct thread *td, struct linux_iopl_args *args)
484 if (args->level < 0 || args->level > 3)
486 if ((error = priv_check(td, PRIV_IO)) != 0)
488 if ((error = securelevel_gt(td->td_ucred, 0)) != 0)
490 td->td_frame->tf_rflags = (td->td_frame->tf_rflags & ~PSL_IOPL) |
491 (args->level * (PSL_IOPL / 3));
497 linux_sigaction(struct thread *td, struct linux_sigaction_args *args)
500 l_sigaction_t act, oact;
503 if (args->nsa != NULL) {
504 error = copyin(args->nsa, &osa, sizeof(l_osigaction_t));
507 act.lsa_handler = osa.lsa_handler;
508 act.lsa_flags = osa.lsa_flags;
509 act.lsa_restorer = osa.lsa_restorer;
510 LINUX_SIGEMPTYSET(act.lsa_mask);
511 act.lsa_mask.__mask = osa.lsa_mask;
514 error = linux_do_sigaction(td, args->sig, args->nsa ? &act : NULL,
515 args->osa ? &oact : NULL);
517 if (args->osa != NULL && !error) {
518 osa.lsa_handler = oact.lsa_handler;
519 osa.lsa_flags = oact.lsa_flags;
520 osa.lsa_restorer = oact.lsa_restorer;
521 osa.lsa_mask = oact.lsa_mask.__mask;
522 error = copyout(&osa, args->osa, sizeof(l_osigaction_t));
529 * Linux has two extra args, restart and oldmask. We don't use these,
530 * but it seems that "restart" is actually a context pointer that
531 * enables the signal to happen with a different register set.
534 linux_sigsuspend(struct thread *td, struct linux_sigsuspend_args *args)
539 LINUX_SIGEMPTYSET(mask);
540 mask.__mask = args->mask;
541 linux_to_bsd_sigset(&mask, &sigmask);
542 return (kern_sigsuspend(td, sigmask));
546 linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap)
552 if (uap->sigsetsize != sizeof(l_sigset_t))
555 error = copyin(uap->newset, &lmask, sizeof(l_sigset_t));
559 linux_to_bsd_sigset(&lmask, &sigmask);
560 return (kern_sigsuspend(td, sigmask));
564 linux_pause(struct thread *td, struct linux_pause_args *args)
566 struct proc *p = td->td_proc;
570 sigmask = td->td_sigmask;
572 return (kern_sigsuspend(td, sigmask));
576 linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap)
582 if (uap->uss != NULL) {
583 error = copyin(uap->uss, &lss, sizeof(l_stack_t));
587 ss.ss_sp = PTRIN(lss.ss_sp);
588 ss.ss_size = lss.ss_size;
589 ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags);
591 error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL,
592 (uap->uoss != NULL) ? &oss : NULL);
593 if (!error && uap->uoss != NULL) {
594 lss.ss_sp = PTROUT(oss.ss_sp);
595 lss.ss_size = oss.ss_size;
596 lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
597 error = copyout(&lss, uap->uoss, sizeof(l_stack_t));
604 linux_gettimeofday(struct thread *td, struct linux_gettimeofday_args *uap)
613 atv32.tv_sec = atv.tv_sec;
614 atv32.tv_usec = atv.tv_usec;
615 error = copyout(&atv32, uap->tp, sizeof(atv32));
617 if (error == 0 && uap->tzp != NULL) {
618 rtz.tz_minuteswest = 0;
620 error = copyout(&rtz, uap->tzp, sizeof(rtz));
626 linux_settimeofday(struct thread *td, struct linux_settimeofday_args *uap)
629 struct timeval atv, *tvp;
630 struct timezone atz, *tzp;
634 error = copyin(uap->tp, &atv32, sizeof(atv32));
637 atv.tv_sec = atv32.tv_sec;
638 atv.tv_usec = atv32.tv_usec;
643 error = copyin(uap->tzp, &atz, sizeof(atz));
649 return (kern_settimeofday(td, tvp, tzp));
653 linux_getrusage(struct thread *td, struct linux_getrusage_args *uap)
658 error = kern_getrusage(td, uap->who, &s);
661 if (uap->rusage != NULL)
662 error = linux_copyout_rusage(&s, uap->rusage);
667 linux_set_thread_area(struct thread *td,
668 struct linux_set_thread_area_args *args)
670 struct l_user_desc info;
671 struct user_segment_descriptor sd;
676 error = copyin(args->desc, &info, sizeof(struct l_user_desc));
681 * Semantics of Linux version: every thread in the system has array
682 * of three TLS descriptors. 1st is GLIBC TLS, 2nd is WINE, 3rd unknown.
683 * This syscall loads one of the selected TLS decriptors with a value
684 * and also loads GDT descriptors 6, 7 and 8 with the content of
685 * the per-thread descriptors.
687 * Semantics of FreeBSD version: I think we can ignore that Linux has
688 * three per-thread descriptors and use just the first one.
689 * The tls_array[] is used only in [gs]et_thread_area() syscalls and
690 * for loading the GDT descriptors. We use just one GDT descriptor
691 * for TLS, so we will load just one.
693 * XXX: This doesn't work when a user space process tries to use more
694 * than one TLS segment. Comment in the Linux source says wine might
699 * GLIBC reads current %gs and call set_thread_area() with it.
700 * We should let GUDATA_SEL and GUGS32_SEL proceed as well because
701 * we use these segments.
703 switch (info.entry_number) {
708 info.entry_number = GUGS32_SEL;
715 * We have to copy out the GDT entry we use.
717 * XXX: What if a user space program does not check the return value
718 * and tries to use 6, 7 or 8?
720 error = copyout(&info, args->desc, sizeof(struct l_user_desc));
724 if (LINUX_LDT_empty(&info)) {
728 a[0] = LINUX_LDT_entry_a(&info);
729 a[1] = LINUX_LDT_entry_b(&info);
732 memcpy(&sd, &a, sizeof(a));
734 pcb->pcb_gsbase = (register_t)info.base_addr;
735 set_pcb_flags(pcb, PCB_32BIT);
736 update_gdt_gsbase(td, info.base_addr);
741 int futex_xchgl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
742 int futex_xchgl_smap(int oparg, uint32_t *uaddr, int *oldval);
743 DEFINE_IFUNC(, int, futex_xchgl, (int, uint32_t *, int *))
746 return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
747 futex_xchgl_smap : futex_xchgl_nosmap);
750 int futex_addl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
751 int futex_addl_smap(int oparg, uint32_t *uaddr, int *oldval);
752 DEFINE_IFUNC(, int, futex_addl, (int, uint32_t *, int *))
755 return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
756 futex_addl_smap : futex_addl_nosmap);
759 int futex_orl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
760 int futex_orl_smap(int oparg, uint32_t *uaddr, int *oldval);
761 DEFINE_IFUNC(, int, futex_orl, (int, uint32_t *, int *))
764 return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
765 futex_orl_smap : futex_orl_nosmap);
768 int futex_andl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
769 int futex_andl_smap(int oparg, uint32_t *uaddr, int *oldval);
770 DEFINE_IFUNC(, int, futex_andl, (int, uint32_t *, int *))
773 return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
774 futex_andl_smap : futex_andl_nosmap);
777 int futex_xorl_nosmap(int oparg, uint32_t *uaddr, int *oldval);
778 int futex_xorl_smap(int oparg, uint32_t *uaddr, int *oldval);
779 DEFINE_IFUNC(, int, futex_xorl, (int, uint32_t *, int *))
782 return ((cpu_stdext_feature & CPUID_STDEXT_SMAP) != 0 ?
783 futex_xorl_smap : futex_xorl_nosmap);