2 * SPDX-License-Identifier: BSD-2-Clause
4 * Copyright (c) 1994-1995 Søren Schmidt
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
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.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/param.h>
33 #include <sys/systm.h>
34 #include <sys/dirent.h>
35 #include <sys/fcntl.h>
37 #include <sys/filedesc.h>
40 #include <sys/selinfo.h>
45 #include <sys/syscallsubr.h>
47 #include <sys/unistd.h>
48 #include <sys/vnode.h>
51 #include <compat/freebsd32/freebsd32_misc.h>
52 #include <machine/../linux32/linux.h>
53 #include <machine/../linux32/linux32_proto.h>
55 #include <machine/../linux/linux.h>
56 #include <machine/../linux/linux_proto.h>
58 #include <compat/linux/linux_misc.h>
59 #include <compat/linux/linux_util.h>
60 #include <compat/linux/linux_file.h>
62 static int linux_common_open(struct thread *, int, const char *, int, int,
64 static int linux_do_accessat(struct thread *t, int, const char *, int, int);
65 static int linux_getdents_error(struct thread *, int, int);
67 static struct bsd_to_linux_bitmap seal_bitmap[] = {
68 BITMAP_1t1_LINUX(F_SEAL_SEAL),
69 BITMAP_1t1_LINUX(F_SEAL_SHRINK),
70 BITMAP_1t1_LINUX(F_SEAL_GROW),
71 BITMAP_1t1_LINUX(F_SEAL_WRITE),
74 #define MFD_HUGETLB_ENTRY(_size) \
76 .bsd_value = MFD_HUGE_##_size, \
77 .linux_value = LINUX_HUGETLB_FLAG_ENCODE_##_size \
79 static struct bsd_to_linux_bitmap mfd_bitmap[] = {
80 BITMAP_1t1_LINUX(MFD_CLOEXEC),
81 BITMAP_1t1_LINUX(MFD_ALLOW_SEALING),
82 BITMAP_1t1_LINUX(MFD_HUGETLB),
83 MFD_HUGETLB_ENTRY(64KB),
84 MFD_HUGETLB_ENTRY(512KB),
85 MFD_HUGETLB_ENTRY(1MB),
86 MFD_HUGETLB_ENTRY(2MB),
87 MFD_HUGETLB_ENTRY(8MB),
88 MFD_HUGETLB_ENTRY(16MB),
89 MFD_HUGETLB_ENTRY(32MB),
90 MFD_HUGETLB_ENTRY(256MB),
91 MFD_HUGETLB_ENTRY(512MB),
92 MFD_HUGETLB_ENTRY(1GB),
93 MFD_HUGETLB_ENTRY(2GB),
94 MFD_HUGETLB_ENTRY(16GB),
96 #undef MFD_HUGETLB_ENTRY
98 #ifdef LINUX_LEGACY_SYSCALLS
100 linux_creat(struct thread *td, struct linux_creat_args *args)
103 return (kern_openat(td, AT_FDCWD, args->path, UIO_USERSPACE,
104 O_WRONLY | O_CREAT | O_TRUNC, args->mode));
109 linux_common_openflags(int l_flags)
114 switch (l_flags & LINUX_O_ACCMODE) {
116 bsd_flags |= O_WRONLY;
122 bsd_flags |= O_RDONLY;
124 if (l_flags & LINUX_O_NDELAY)
125 bsd_flags |= O_NONBLOCK;
126 if (l_flags & LINUX_O_APPEND)
127 bsd_flags |= O_APPEND;
128 if (l_flags & LINUX_O_SYNC)
129 bsd_flags |= O_FSYNC;
130 if (l_flags & LINUX_O_CLOEXEC)
131 bsd_flags |= O_CLOEXEC;
132 if (l_flags & LINUX_O_NONBLOCK)
133 bsd_flags |= O_NONBLOCK;
134 if (l_flags & LINUX_O_ASYNC)
135 bsd_flags |= O_ASYNC;
136 if (l_flags & LINUX_O_CREAT)
137 bsd_flags |= O_CREAT;
138 if (l_flags & LINUX_O_TRUNC)
139 bsd_flags |= O_TRUNC;
140 if (l_flags & LINUX_O_EXCL)
142 if (l_flags & LINUX_O_NOCTTY)
143 bsd_flags |= O_NOCTTY;
144 if (l_flags & LINUX_O_DIRECT)
145 bsd_flags |= O_DIRECT;
146 if (l_flags & LINUX_O_NOFOLLOW)
147 bsd_flags |= O_NOFOLLOW;
148 if (l_flags & LINUX_O_DIRECTORY)
149 bsd_flags |= O_DIRECTORY;
150 if (l_flags & LINUX_O_PATH)
152 /* XXX LINUX_O_NOATIME: unable to be easily implemented. */
157 linux_common_open(struct thread *td, int dirfd, const char *path, int l_flags,
158 int mode, enum uio_seg seg)
160 struct proc *p = td->td_proc;
163 int bsd_flags, error;
165 bsd_flags = linux_common_openflags(l_flags);
166 error = kern_openat(td, dirfd, path, seg, bsd_flags, mode);
172 if (p->p_flag & P_CONTROLT)
174 if (bsd_flags & O_NOCTTY)
178 * XXX In between kern_openat() and fget(), another process
179 * having the same filedesc could use that fd without
182 fd = td->td_retval[0];
183 if (fget(td, fd, &cap_ioctl_rights, &fp) == 0) {
184 if (fp->f_type != DTYPE_VNODE) {
188 sx_slock(&proctree_lock);
190 if (SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
192 sx_sunlock(&proctree_lock);
193 /* XXXPJD: Verify if TIOCSCTTY is allowed. */
194 (void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0,
198 sx_sunlock(&proctree_lock);
208 linux_openat(struct thread *td, struct linux_openat_args *args)
212 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
213 return (linux_common_open(td, dfd, args->filename, args->flags,
214 args->mode, UIO_USERSPACE));
217 #ifdef LINUX_LEGACY_SYSCALLS
219 linux_open(struct thread *td, struct linux_open_args *args)
222 return (linux_common_open(td, AT_FDCWD, args->path, args->flags,
223 args->mode, UIO_USERSPACE));
228 linux_name_to_handle_at(struct thread *td,
229 struct linux_name_to_handle_at_args *args)
231 static const l_int valid_flags = (LINUX_AT_SYMLINK_FOLLOW |
232 LINUX_AT_EMPTY_PATH);
233 static const l_uint fh_size = sizeof(fhandle_t);
238 int error, fd, bsd_flags;
240 if (args->flags & ~valid_flags)
244 if (fd == LINUX_AT_FDCWD)
248 if (!(args->flags & LINUX_AT_SYMLINK_FOLLOW))
249 bsd_flags |= AT_SYMLINK_NOFOLLOW;
250 if ((args->flags & LINUX_AT_EMPTY_PATH) != 0)
251 bsd_flags |= AT_EMPTY_PATH;
253 error = kern_getfhat(td, bsd_flags, fd, args->name,
254 UIO_USERSPACE, &fh, UIO_SYSSPACE);
258 /* Emit mount_id -- required before EOVERFLOW case. */
259 mount_id = (fh.fh_fsid.val[0] ^ fh.fh_fsid.val[1]);
260 error = copyout(&mount_id, args->mnt_id, sizeof(mount_id));
264 /* Check if there is room for handle. */
265 error = copyin(&args->handle->handle_bytes, &fh_bytes,
270 if (fh_bytes < fh_size) {
271 error = copyout(&fh_size, &args->handle->handle_bytes,
281 * We don't use handle_type for anything yet, but initialize a known
284 error = copyout(&mount_id, &args->handle->handle_type,
289 error = copyout(&fh, &args->handle->f_handle,
295 linux_open_by_handle_at(struct thread *td,
296 struct linux_open_by_handle_at_args *args)
299 int bsd_flags, error;
301 error = copyin(&args->handle->handle_bytes, &fh_bytes,
306 if (fh_bytes < sizeof(fhandle_t))
309 bsd_flags = linux_common_openflags(args->flags);
310 return (kern_fhopen(td, (void *)&args->handle->f_handle, bsd_flags));
314 linux_lseek(struct thread *td, struct linux_lseek_args *args)
317 return (kern_lseek(td, args->fdes, args->off, args->whence));
320 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
322 linux_llseek(struct thread *td, struct linux_llseek_args *args)
327 off = (args->olow) | (((off_t) args->ohigh) << 32);
329 error = kern_lseek(td, args->fd, off, args->whence);
333 error = copyout(td->td_retval, args->res, sizeof(off_t));
337 td->td_retval[0] = 0;
340 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
343 * Note that linux_getdents(2) and linux_getdents64(2) have the same
344 * arguments. They only differ in the definition of struct dirent they
346 * Note that linux_readdir(2) is a special case of linux_getdents(2)
347 * where count is always equals 1, meaning that the buffer is one
348 * dirent-structure in size and that the code can't handle more anyway.
349 * Note that linux_readdir(2) can't be implemented by means of linux_getdents(2)
350 * as in case when the *dent buffer size is equal to 1 linux_getdents(2) will
355 linux_getdents_error(struct thread *td, int fd, int err)
361 /* Linux return ENOTDIR in case when fd is not a directory. */
362 error = getvnode(td, fd, &cap_read_rights, &fp);
366 if (vp->v_type != VDIR) {
378 char d_name[LINUX_NAME_MAX + 1];
386 char d_name[LINUX_NAME_MAX + 1];
390 * Linux uses the last byte in the dirent buffer to store d_type,
391 * at least glibc-2.7 requires it. That is why l_dirent is padded with 2 bytes.
393 #define LINUX_RECLEN(namlen) \
394 roundup(offsetof(struct l_dirent, d_name) + (namlen) + 2, sizeof(l_ulong))
396 #define LINUX_RECLEN64(namlen) \
397 roundup(offsetof(struct l_dirent64, d_name) + (namlen) + 1, \
400 #ifdef LINUX_LEGACY_SYSCALLS
402 linux_getdents(struct thread *td, struct linux_getdents_args *args)
405 caddr_t inp, buf; /* BSD-format */
406 int len, reclen; /* BSD-format */
407 caddr_t outp; /* Linux-format */
408 int resid, linuxreclen; /* Linux-format */
409 caddr_t lbuf; /* Linux-format */
411 struct l_dirent *linux_dirent;
415 buflen = min(args->count, MAXBSIZE);
416 buf = malloc(buflen, M_TEMP, M_WAITOK);
418 error = kern_getdirentries(td, args->fd, buf, buflen,
419 &base, NULL, UIO_SYSSPACE);
421 error = linux_getdents_error(td, args->fd, error);
425 lbuf = malloc(LINUX_RECLEN(LINUX_NAME_MAX), M_TEMP, M_WAITOK | M_ZERO);
427 len = td->td_retval[0];
429 outp = (caddr_t)args->dent;
434 bdp = (struct dirent *) inp;
435 reclen = bdp->d_reclen;
436 linuxreclen = LINUX_RECLEN(bdp->d_namlen);
438 * No more space in the user supplied dirent buffer.
441 if (resid < linuxreclen) {
446 linux_dirent = (struct l_dirent*)lbuf;
447 linux_dirent->d_ino = bdp->d_fileno;
448 linux_dirent->d_off = bdp->d_off;
449 linux_dirent->d_reclen = linuxreclen;
451 * Copy d_type to last byte of l_dirent buffer
453 lbuf[linuxreclen - 1] = bdp->d_type;
454 strlcpy(linux_dirent->d_name, bdp->d_name,
455 linuxreclen - offsetof(struct l_dirent, d_name)-1);
456 error = copyout(linux_dirent, outp, linuxreclen);
464 retval += linuxreclen;
466 resid -= linuxreclen;
468 td->td_retval[0] = retval;
479 linux_getdents64(struct thread *td, struct linux_getdents64_args *args)
482 caddr_t inp, buf; /* BSD-format */
483 int len, reclen; /* BSD-format */
484 caddr_t outp; /* Linux-format */
485 int resid, linuxreclen; /* Linux-format */
486 caddr_t lbuf; /* Linux-format */
488 struct l_dirent64 *linux_dirent64;
492 buflen = min(args->count, MAXBSIZE);
493 buf = malloc(buflen, M_TEMP, M_WAITOK);
495 error = kern_getdirentries(td, args->fd, buf, buflen,
496 &base, NULL, UIO_SYSSPACE);
498 error = linux_getdents_error(td, args->fd, error);
502 lbuf = malloc(LINUX_RECLEN64(LINUX_NAME_MAX), M_TEMP, M_WAITOK | M_ZERO);
504 len = td->td_retval[0];
506 outp = (caddr_t)args->dirent;
511 bdp = (struct dirent *) inp;
512 reclen = bdp->d_reclen;
513 linuxreclen = LINUX_RECLEN64(bdp->d_namlen);
515 * No more space in the user supplied dirent buffer.
518 if (resid < linuxreclen) {
523 linux_dirent64 = (struct l_dirent64*)lbuf;
524 linux_dirent64->d_ino = bdp->d_fileno;
525 linux_dirent64->d_off = bdp->d_off;
526 linux_dirent64->d_reclen = linuxreclen;
527 linux_dirent64->d_type = bdp->d_type;
528 strlcpy(linux_dirent64->d_name, bdp->d_name,
529 linuxreclen - offsetof(struct l_dirent64, d_name));
530 error = copyout(linux_dirent64, outp, linuxreclen);
538 retval += linuxreclen;
540 resid -= linuxreclen;
542 td->td_retval[0] = retval;
551 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
553 linux_readdir(struct thread *td, struct linux_readdir_args *args)
556 caddr_t buf; /* BSD-format */
557 int linuxreclen; /* Linux-format */
558 caddr_t lbuf; /* Linux-format */
560 struct l_dirent *linux_dirent;
563 buflen = LINUX_RECLEN(LINUX_NAME_MAX);
564 buf = malloc(buflen, M_TEMP, M_WAITOK);
566 error = kern_getdirentries(td, args->fd, buf, buflen,
567 &base, NULL, UIO_SYSSPACE);
569 error = linux_getdents_error(td, args->fd, error);
572 if (td->td_retval[0] == 0)
575 lbuf = malloc(LINUX_RECLEN(LINUX_NAME_MAX), M_TEMP, M_WAITOK | M_ZERO);
577 bdp = (struct dirent *) buf;
578 linuxreclen = LINUX_RECLEN(bdp->d_namlen);
580 linux_dirent = (struct l_dirent*)lbuf;
581 linux_dirent->d_ino = bdp->d_fileno;
582 linux_dirent->d_off = bdp->d_off;
583 linux_dirent->d_reclen = bdp->d_namlen;
584 strlcpy(linux_dirent->d_name, bdp->d_name,
585 linuxreclen - offsetof(struct l_dirent, d_name));
586 error = copyout(linux_dirent, args->dent, linuxreclen);
588 td->td_retval[0] = linuxreclen;
595 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
598 * These exist mainly for hooks for doing /compat/linux translation.
601 #ifdef LINUX_LEGACY_SYSCALLS
603 linux_access(struct thread *td, struct linux_access_args *args)
606 /* Linux convention. */
607 if (args->amode & ~(F_OK | X_OK | W_OK | R_OK))
610 return (kern_accessat(td, AT_FDCWD, args->path, UIO_USERSPACE, 0,
616 linux_do_accessat(struct thread *td, int ldfd, const char *filename,
617 int amode, int flags)
621 /* Linux convention. */
622 if (amode & ~(F_OK | X_OK | W_OK | R_OK))
625 dfd = (ldfd == LINUX_AT_FDCWD) ? AT_FDCWD : ldfd;
626 return (kern_accessat(td, dfd, filename, UIO_USERSPACE, flags, amode));
630 linux_faccessat(struct thread *td, struct linux_faccessat_args *args)
633 return (linux_do_accessat(td, args->dfd, args->filename, args->amode,
638 linux_faccessat2(struct thread *td, struct linux_faccessat2_args *args)
640 int flags, unsupported;
642 /* XXX. AT_SYMLINK_NOFOLLOW is not supported by kern_accessat */
643 unsupported = args->flags & ~(LINUX_AT_EACCESS | LINUX_AT_EMPTY_PATH);
644 if (unsupported != 0) {
645 linux_msg(td, "faccessat2 unsupported flag 0x%x", unsupported);
649 flags = (args->flags & LINUX_AT_EACCESS) == 0 ? 0 :
651 flags |= (args->flags & LINUX_AT_EMPTY_PATH) == 0 ? 0 :
653 return (linux_do_accessat(td, args->dfd, args->filename, args->amode,
658 #ifdef LINUX_LEGACY_SYSCALLS
660 linux_unlink(struct thread *td, struct linux_unlink_args *args)
665 error = kern_funlinkat(td, AT_FDCWD, args->path, FD_NONE,
666 UIO_USERSPACE, 0, 0);
667 if (error == EPERM) {
668 /* Introduce POSIX noncompliant behaviour of Linux */
669 if (kern_statat(td, 0, AT_FDCWD, args->path,
670 UIO_USERSPACE, &st) == 0) {
671 if (S_ISDIR(st.st_mode))
681 linux_unlinkat_impl(struct thread *td, enum uio_seg pathseg, const char *path,
682 int dfd, struct linux_unlinkat_args *args)
687 if (args->flag & LINUX_AT_REMOVEDIR)
688 error = kern_frmdirat(td, dfd, path, FD_NONE, pathseg, 0);
690 error = kern_funlinkat(td, dfd, path, FD_NONE, pathseg, 0, 0);
691 if (error == EPERM && !(args->flag & LINUX_AT_REMOVEDIR)) {
692 /* Introduce POSIX noncompliant behaviour of Linux */
693 if (kern_statat(td, AT_SYMLINK_NOFOLLOW, dfd, path,
694 pathseg, &st) == 0 && S_ISDIR(st.st_mode))
701 linux_unlinkat(struct thread *td, struct linux_unlinkat_args *args)
705 if (args->flag & ~LINUX_AT_REMOVEDIR)
707 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
708 return (linux_unlinkat_impl(td, UIO_USERSPACE, args->pathname,
713 linux_chdir(struct thread *td, struct linux_chdir_args *args)
716 return (kern_chdir(td, args->path, UIO_USERSPACE));
719 #ifdef LINUX_LEGACY_SYSCALLS
721 linux_chmod(struct thread *td, struct linux_chmod_args *args)
724 return (kern_fchmodat(td, AT_FDCWD, args->path, UIO_USERSPACE,
730 linux_fchmodat(struct thread *td, struct linux_fchmodat_args *args)
734 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
735 return (kern_fchmodat(td, dfd, args->filename, UIO_USERSPACE,
739 #ifdef LINUX_LEGACY_SYSCALLS
741 linux_mkdir(struct thread *td, struct linux_mkdir_args *args)
744 return (kern_mkdirat(td, AT_FDCWD, args->path, UIO_USERSPACE, args->mode));
749 linux_mkdirat(struct thread *td, struct linux_mkdirat_args *args)
753 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
754 return (kern_mkdirat(td, dfd, args->pathname, UIO_USERSPACE, args->mode));
757 #ifdef LINUX_LEGACY_SYSCALLS
759 linux_rmdir(struct thread *td, struct linux_rmdir_args *args)
762 return (kern_frmdirat(td, AT_FDCWD, args->path, FD_NONE,
767 linux_rename(struct thread *td, struct linux_rename_args *args)
770 return (kern_renameat(td, AT_FDCWD, args->from, AT_FDCWD,
771 args->to, UIO_USERSPACE));
776 linux_renameat(struct thread *td, struct linux_renameat_args *args)
778 struct linux_renameat2_args renameat2_args = {
779 .olddfd = args->olddfd,
780 .oldname = args->oldname,
781 .newdfd = args->newdfd,
782 .newname = args->newname,
786 return (linux_renameat2(td, &renameat2_args));
790 linux_renameat2(struct thread *td, struct linux_renameat2_args *args)
794 if (args->flags != 0) {
795 if (args->flags & ~(LINUX_RENAME_EXCHANGE |
796 LINUX_RENAME_NOREPLACE | LINUX_RENAME_WHITEOUT))
798 if (args->flags & LINUX_RENAME_EXCHANGE &&
799 args->flags & (LINUX_RENAME_NOREPLACE |
800 LINUX_RENAME_WHITEOUT))
804 * This spams the console on Ubuntu Focal.
806 * What's needed here is a general mechanism to let users know
807 * about missing features without hogging the system.
809 linux_msg(td, "renameat2 unsupported flags 0x%x",
815 olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
816 newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
817 return (kern_renameat(td, olddfd, args->oldname, newdfd,
818 args->newname, UIO_USERSPACE));
821 #ifdef LINUX_LEGACY_SYSCALLS
823 linux_symlink(struct thread *td, struct linux_symlink_args *args)
826 return (kern_symlinkat(td, args->path, AT_FDCWD, args->to,
832 linux_symlinkat(struct thread *td, struct linux_symlinkat_args *args)
836 dfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
837 return (kern_symlinkat(td, args->oldname, dfd, args->newname,
841 #ifdef LINUX_LEGACY_SYSCALLS
843 linux_readlink(struct thread *td, struct linux_readlink_args *args)
846 if (args->count <= 0)
849 return (kern_readlinkat(td, AT_FDCWD, args->name, UIO_USERSPACE,
850 args->buf, UIO_USERSPACE, args->count));
855 linux_readlinkat(struct thread *td, struct linux_readlinkat_args *args)
859 if (args->bufsiz <= 0)
862 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
863 return (kern_readlinkat(td, dfd, args->path, UIO_USERSPACE,
864 args->buf, UIO_USERSPACE, args->bufsiz));
868 linux_truncate(struct thread *td, struct linux_truncate_args *args)
871 return (kern_truncate(td, args->path, UIO_USERSPACE, args->length));
874 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
876 linux_truncate64(struct thread *td, struct linux_truncate64_args *args)
880 #if defined(__amd64__) && defined(COMPAT_LINUX32)
881 length = PAIR32TO64(off_t, args->length);
883 length = args->length;
886 return (kern_truncate(td, args->path, UIO_USERSPACE, length));
888 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
891 linux_ftruncate(struct thread *td, struct linux_ftruncate_args *args)
894 return (kern_ftruncate(td, args->fd, args->length));
897 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
899 linux_ftruncate64(struct thread *td, struct linux_ftruncate64_args *args)
903 #if defined(__amd64__) && defined(COMPAT_LINUX32)
904 length = PAIR32TO64(off_t, args->length);
906 length = args->length;
909 return (kern_ftruncate(td, args->fd, length));
913 #ifdef LINUX_LEGACY_SYSCALLS
915 linux_link(struct thread *td, struct linux_link_args *args)
918 return (kern_linkat(td, AT_FDCWD, AT_FDCWD, args->path, args->to,
919 UIO_USERSPACE, AT_SYMLINK_FOLLOW));
924 linux_linkat(struct thread *td, struct linux_linkat_args *args)
926 int olddfd, newdfd, flag;
928 if (args->flag & ~(LINUX_AT_SYMLINK_FOLLOW | LINUX_AT_EMPTY_PATH))
931 flag = (args->flag & LINUX_AT_SYMLINK_FOLLOW) != 0 ? AT_SYMLINK_FOLLOW :
933 flag |= (args->flag & LINUX_AT_EMPTY_PATH) != 0 ? AT_EMPTY_PATH : 0;
935 olddfd = (args->olddfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->olddfd;
936 newdfd = (args->newdfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->newdfd;
937 return (kern_linkat(td, olddfd, newdfd, args->oldname,
938 args->newname, UIO_USERSPACE, flag));
942 linux_fdatasync(struct thread *td, struct linux_fdatasync_args *uap)
945 return (kern_fsync(td, uap->fd, false));
949 linux_sync_file_range(struct thread *td, struct linux_sync_file_range_args *uap)
951 off_t nbytes, offset;
953 #if defined(__amd64__) && defined(COMPAT_LINUX32)
954 nbytes = PAIR32TO64(off_t, uap->nbytes);
955 offset = PAIR32TO64(off_t, uap->offset);
957 nbytes = uap->nbytes;
958 offset = uap->offset;
961 if (offset < 0 || nbytes < 0 ||
962 (uap->flags & ~(LINUX_SYNC_FILE_RANGE_WAIT_BEFORE |
963 LINUX_SYNC_FILE_RANGE_WRITE |
964 LINUX_SYNC_FILE_RANGE_WAIT_AFTER)) != 0) {
968 return (kern_fsync(td, uap->fd, false));
972 linux_pread(struct thread *td, struct linux_pread_args *uap)
978 #if defined(__amd64__) && defined(COMPAT_LINUX32)
979 offset = PAIR32TO64(off_t, uap->offset);
981 offset = uap->offset;
984 error = kern_pread(td, uap->fd, uap->buf, uap->nbyte, offset);
986 /* This seems to violate POSIX but Linux does it. */
987 error = fgetvp(td, uap->fd, &cap_pread_rights, &vp);
990 if (vp->v_type == VDIR)
998 linux_pwrite(struct thread *td, struct linux_pwrite_args *uap)
1002 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1003 offset = PAIR32TO64(off_t, uap->offset);
1005 offset = uap->offset;
1008 return (kern_pwrite(td, uap->fd, uap->buf, uap->nbyte, offset));
1011 #define HALF_LONG_BITS ((sizeof(l_long) * NBBY / 2))
1014 pos_from_hilo(unsigned long high, unsigned long low)
1017 return (((off_t)high << HALF_LONG_BITS) << HALF_LONG_BITS) | low;
1021 linux_preadv(struct thread *td, struct linux_preadv_args *uap)
1028 * According http://man7.org/linux/man-pages/man2/preadv.2.html#NOTES
1029 * pos_l and pos_h, respectively, contain the
1030 * low order and high order 32 bits of offset.
1032 offset = pos_from_hilo(uap->pos_h, uap->pos_l);
1035 #ifdef COMPAT_LINUX32
1036 error = linux32_copyinuio(PTRIN(uap->vec), uap->vlen, &auio);
1038 error = copyinuio(uap->vec, uap->vlen, &auio);
1042 error = kern_preadv(td, uap->fd, auio, offset);
1048 linux_pwritev(struct thread *td, struct linux_pwritev_args *uap)
1055 * According http://man7.org/linux/man-pages/man2/pwritev.2.html#NOTES
1056 * pos_l and pos_h, respectively, contain the
1057 * low order and high order 32 bits of offset.
1059 offset = pos_from_hilo(uap->pos_h, uap->pos_l);
1062 #ifdef COMPAT_LINUX32
1063 error = linux32_copyinuio(PTRIN(uap->vec), uap->vlen, &auio);
1065 error = copyinuio(uap->vec, uap->vlen, &auio);
1069 error = kern_pwritev(td, uap->fd, auio, offset);
1075 linux_mount(struct thread *td, struct linux_mount_args *args)
1077 struct mntarg *ma = NULL;
1078 char *fstypename, *mntonname, *mntfromname, *data;
1081 fstypename = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1082 mntonname = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1083 mntfromname = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1085 error = copyinstr(args->filesystemtype, fstypename, MNAMELEN - 1,
1089 if (args->specialfile != NULL) {
1090 error = copyinstr(args->specialfile, mntfromname, MNAMELEN - 1, NULL);
1094 mntfromname[0] = '\0';
1096 error = copyinstr(args->dir, mntonname, MNAMELEN - 1, NULL);
1100 if (strcmp(fstypename, "ext2") == 0) {
1101 strcpy(fstypename, "ext2fs");
1102 } else if (strcmp(fstypename, "proc") == 0) {
1103 strcpy(fstypename, "linprocfs");
1104 } else if (strcmp(fstypename, "vfat") == 0) {
1105 strcpy(fstypename, "msdosfs");
1106 } else if (strcmp(fstypename, "fuse") == 0 ||
1107 strncmp(fstypename, "fuse.", 5) == 0) {
1108 char *fuse_options, *fuse_option, *fuse_name;
1110 strcpy(mntfromname, "/dev/fuse");
1111 strcpy(fstypename, "fusefs");
1112 data = malloc(MNAMELEN, M_TEMP, M_WAITOK);
1113 error = copyinstr(args->data, data, MNAMELEN - 1, NULL);
1117 fuse_options = data;
1118 while ((fuse_option = strsep(&fuse_options, ",")) != NULL) {
1119 fuse_name = strsep(&fuse_option, "=");
1120 if (fuse_name == NULL || fuse_option == NULL)
1122 ma = mount_arg(ma, fuse_name, fuse_option, -1);
1126 * The FUSE server uses Linux errno values instead of FreeBSD
1127 * ones; add a flag to tell fuse(4) to do errno translation.
1129 ma = mount_arg(ma, "linux_errnos", "1", -1);
1135 * Linux SYNC flag is not included; the closest equivalent
1136 * FreeBSD has is !ASYNC, which is our default.
1138 if (args->rwflag & LINUX_MS_RDONLY)
1139 fsflags |= MNT_RDONLY;
1140 if (args->rwflag & LINUX_MS_NOSUID)
1141 fsflags |= MNT_NOSUID;
1142 if (args->rwflag & LINUX_MS_NOEXEC)
1143 fsflags |= MNT_NOEXEC;
1144 if (args->rwflag & LINUX_MS_REMOUNT)
1145 fsflags |= MNT_UPDATE;
1147 ma = mount_arg(ma, "fstype", fstypename, -1);
1148 ma = mount_arg(ma, "fspath", mntonname, -1);
1149 ma = mount_arg(ma, "from", mntfromname, -1);
1150 error = kernel_mount(ma, fsflags);
1152 free(fstypename, M_TEMP);
1153 free(mntonname, M_TEMP);
1154 free(mntfromname, M_TEMP);
1159 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1161 linux_oldumount(struct thread *td, struct linux_oldumount_args *args)
1164 return (kern_unmount(td, args->path, 0));
1166 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1168 #ifdef LINUX_LEGACY_SYSCALLS
1170 linux_umount(struct thread *td, struct linux_umount_args *args)
1175 if ((args->flags & LINUX_MNT_FORCE) != 0) {
1176 args->flags &= ~LINUX_MNT_FORCE;
1179 if (args->flags != 0) {
1180 linux_msg(td, "unsupported umount2 flags %#x", args->flags);
1184 return (kern_unmount(td, args->path, flags));
1189 * fcntl family of syscalls
1199 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1205 linux_to_bsd_flock(struct l_flock *linux_flock, struct flock *bsd_flock)
1207 switch (linux_flock->l_type) {
1209 bsd_flock->l_type = F_RDLCK;
1212 bsd_flock->l_type = F_WRLCK;
1215 bsd_flock->l_type = F_UNLCK;
1218 bsd_flock->l_type = -1;
1221 bsd_flock->l_whence = linux_flock->l_whence;
1222 bsd_flock->l_start = (off_t)linux_flock->l_start;
1223 bsd_flock->l_len = (off_t)linux_flock->l_len;
1224 bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1225 bsd_flock->l_sysid = 0;
1229 bsd_to_linux_flock(struct flock *bsd_flock, struct l_flock *linux_flock)
1231 switch (bsd_flock->l_type) {
1233 linux_flock->l_type = LINUX_F_RDLCK;
1236 linux_flock->l_type = LINUX_F_WRLCK;
1239 linux_flock->l_type = LINUX_F_UNLCK;
1242 linux_flock->l_whence = bsd_flock->l_whence;
1243 linux_flock->l_start = (l_off_t)bsd_flock->l_start;
1244 linux_flock->l_len = (l_off_t)bsd_flock->l_len;
1245 linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1248 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1256 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1262 linux_to_bsd_flock64(struct l_flock64 *linux_flock, struct flock *bsd_flock)
1264 switch (linux_flock->l_type) {
1266 bsd_flock->l_type = F_RDLCK;
1269 bsd_flock->l_type = F_WRLCK;
1272 bsd_flock->l_type = F_UNLCK;
1275 bsd_flock->l_type = -1;
1278 bsd_flock->l_whence = linux_flock->l_whence;
1279 bsd_flock->l_start = (off_t)linux_flock->l_start;
1280 bsd_flock->l_len = (off_t)linux_flock->l_len;
1281 bsd_flock->l_pid = (pid_t)linux_flock->l_pid;
1282 bsd_flock->l_sysid = 0;
1286 bsd_to_linux_flock64(struct flock *bsd_flock, struct l_flock64 *linux_flock)
1288 switch (bsd_flock->l_type) {
1290 linux_flock->l_type = LINUX_F_RDLCK;
1293 linux_flock->l_type = LINUX_F_WRLCK;
1296 linux_flock->l_type = LINUX_F_UNLCK;
1299 linux_flock->l_whence = bsd_flock->l_whence;
1300 linux_flock->l_start = (l_loff_t)bsd_flock->l_start;
1301 linux_flock->l_len = (l_loff_t)bsd_flock->l_len;
1302 linux_flock->l_pid = (l_pid_t)bsd_flock->l_pid;
1304 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1307 fcntl_common(struct thread *td, struct linux_fcntl_args *args)
1309 struct l_flock linux_flock;
1310 struct flock bsd_flock;
1316 switch (args->cmd) {
1318 return (kern_fcntl(td, args->fd, F_DUPFD, args->arg));
1321 return (kern_fcntl(td, args->fd, F_GETFD, 0));
1324 return (kern_fcntl(td, args->fd, F_SETFD, args->arg));
1327 error = kern_fcntl(td, args->fd, F_GETFL, 0);
1328 result = td->td_retval[0];
1329 td->td_retval[0] = 0;
1330 if (result & O_RDONLY)
1331 td->td_retval[0] |= LINUX_O_RDONLY;
1332 if (result & O_WRONLY)
1333 td->td_retval[0] |= LINUX_O_WRONLY;
1334 if (result & O_RDWR)
1335 td->td_retval[0] |= LINUX_O_RDWR;
1336 if (result & O_NDELAY)
1337 td->td_retval[0] |= LINUX_O_NONBLOCK;
1338 if (result & O_APPEND)
1339 td->td_retval[0] |= LINUX_O_APPEND;
1340 if (result & O_FSYNC)
1341 td->td_retval[0] |= LINUX_O_SYNC;
1342 if (result & O_ASYNC)
1343 td->td_retval[0] |= LINUX_O_ASYNC;
1344 #ifdef LINUX_O_NOFOLLOW
1345 if (result & O_NOFOLLOW)
1346 td->td_retval[0] |= LINUX_O_NOFOLLOW;
1348 #ifdef LINUX_O_DIRECT
1349 if (result & O_DIRECT)
1350 td->td_retval[0] |= LINUX_O_DIRECT;
1356 if (args->arg & LINUX_O_NDELAY)
1358 if (args->arg & LINUX_O_APPEND)
1360 if (args->arg & LINUX_O_SYNC)
1362 if (args->arg & LINUX_O_ASYNC)
1364 #ifdef LINUX_O_NOFOLLOW
1365 if (args->arg & LINUX_O_NOFOLLOW)
1368 #ifdef LINUX_O_DIRECT
1369 if (args->arg & LINUX_O_DIRECT)
1372 return (kern_fcntl(td, args->fd, F_SETFL, arg));
1375 error = copyin((void *)args->arg, &linux_flock,
1376 sizeof(linux_flock));
1379 linux_to_bsd_flock(&linux_flock, &bsd_flock);
1380 error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1383 bsd_to_linux_flock(&bsd_flock, &linux_flock);
1384 return (copyout(&linux_flock, (void *)args->arg,
1385 sizeof(linux_flock)));
1388 error = copyin((void *)args->arg, &linux_flock,
1389 sizeof(linux_flock));
1392 linux_to_bsd_flock(&linux_flock, &bsd_flock);
1393 return (kern_fcntl(td, args->fd, F_SETLK,
1394 (intptr_t)&bsd_flock));
1396 case LINUX_F_SETLKW:
1397 error = copyin((void *)args->arg, &linux_flock,
1398 sizeof(linux_flock));
1401 linux_to_bsd_flock(&linux_flock, &bsd_flock);
1402 return (kern_fcntl(td, args->fd, F_SETLKW,
1403 (intptr_t)&bsd_flock));
1405 case LINUX_F_GETOWN:
1406 return (kern_fcntl(td, args->fd, F_GETOWN, 0));
1408 case LINUX_F_SETOWN:
1410 * XXX some Linux applications depend on F_SETOWN having no
1411 * significant effect for pipes (SIGIO is not delivered for
1412 * pipes under Linux-2.2.35 at least).
1414 error = fget(td, args->fd,
1415 &cap_fcntl_rights, &fp);
1418 if (fp->f_type == DTYPE_PIPE) {
1424 return (kern_fcntl(td, args->fd, F_SETOWN, args->arg));
1426 case LINUX_F_DUPFD_CLOEXEC:
1427 return (kern_fcntl(td, args->fd, F_DUPFD_CLOEXEC, args->arg));
1429 * Our F_SEAL_* values match Linux one for maximum compatibility. So we
1430 * only needed to account for different values for fcntl(2) commands.
1432 case LINUX_F_GET_SEALS:
1433 error = kern_fcntl(td, args->fd, F_GET_SEALS, 0);
1436 td->td_retval[0] = bsd_to_linux_bits(td->td_retval[0],
1440 case LINUX_F_ADD_SEALS:
1441 return (kern_fcntl(td, args->fd, F_ADD_SEALS,
1442 linux_to_bsd_bits(args->arg, seal_bitmap, 0)));
1444 case LINUX_F_GETPIPE_SZ:
1445 error = fget(td, args->fd,
1446 &cap_fcntl_rights, &fp);
1449 if (fp->f_type != DTYPE_PIPE) {
1454 td->td_retval[0] = fpipe->pipe_buffer.size;
1459 linux_msg(td, "unsupported fcntl cmd %d", args->cmd);
1465 linux_fcntl(struct thread *td, struct linux_fcntl_args *args)
1468 return (fcntl_common(td, args));
1471 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1473 linux_fcntl64(struct thread *td, struct linux_fcntl64_args *args)
1475 struct l_flock64 linux_flock;
1476 struct flock bsd_flock;
1477 struct linux_fcntl_args fcntl_args;
1480 switch (args->cmd) {
1481 case LINUX_F_GETLK64:
1482 error = copyin((void *)args->arg, &linux_flock,
1483 sizeof(linux_flock));
1486 linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1487 error = kern_fcntl(td, args->fd, F_GETLK, (intptr_t)&bsd_flock);
1490 bsd_to_linux_flock64(&bsd_flock, &linux_flock);
1491 return (copyout(&linux_flock, (void *)args->arg,
1492 sizeof(linux_flock)));
1494 case LINUX_F_SETLK64:
1495 error = copyin((void *)args->arg, &linux_flock,
1496 sizeof(linux_flock));
1499 linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1500 return (kern_fcntl(td, args->fd, F_SETLK,
1501 (intptr_t)&bsd_flock));
1503 case LINUX_F_SETLKW64:
1504 error = copyin((void *)args->arg, &linux_flock,
1505 sizeof(linux_flock));
1508 linux_to_bsd_flock64(&linux_flock, &bsd_flock);
1509 return (kern_fcntl(td, args->fd, F_SETLKW,
1510 (intptr_t)&bsd_flock));
1513 fcntl_args.fd = args->fd;
1514 fcntl_args.cmd = args->cmd;
1515 fcntl_args.arg = args->arg;
1516 return (fcntl_common(td, &fcntl_args));
1518 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1520 #ifdef LINUX_LEGACY_SYSCALLS
1522 linux_chown(struct thread *td, struct linux_chown_args *args)
1525 return (kern_fchownat(td, AT_FDCWD, args->path, UIO_USERSPACE,
1526 args->uid, args->gid, 0));
1531 linux_fchownat(struct thread *td, struct linux_fchownat_args *args)
1533 int dfd, flag, unsupported;
1535 unsupported = args->flag & ~(LINUX_AT_SYMLINK_NOFOLLOW | LINUX_AT_EMPTY_PATH);
1536 if (unsupported != 0) {
1537 linux_msg(td, "fchownat unsupported flag 0x%x", unsupported);
1541 flag = (args->flag & LINUX_AT_SYMLINK_NOFOLLOW) == 0 ? 0 :
1542 AT_SYMLINK_NOFOLLOW;
1543 flag |= (args->flag & LINUX_AT_EMPTY_PATH) == 0 ? 0 :
1546 dfd = (args->dfd == LINUX_AT_FDCWD) ? AT_FDCWD : args->dfd;
1547 return (kern_fchownat(td, dfd, args->filename, UIO_USERSPACE,
1548 args->uid, args->gid, flag));
1551 #ifdef LINUX_LEGACY_SYSCALLS
1553 linux_lchown(struct thread *td, struct linux_lchown_args *args)
1556 return (kern_fchownat(td, AT_FDCWD, args->path, UIO_USERSPACE, args->uid,
1557 args->gid, AT_SYMLINK_NOFOLLOW));
1562 convert_fadvice(int advice)
1565 case LINUX_POSIX_FADV_NORMAL:
1566 return (POSIX_FADV_NORMAL);
1567 case LINUX_POSIX_FADV_RANDOM:
1568 return (POSIX_FADV_RANDOM);
1569 case LINUX_POSIX_FADV_SEQUENTIAL:
1570 return (POSIX_FADV_SEQUENTIAL);
1571 case LINUX_POSIX_FADV_WILLNEED:
1572 return (POSIX_FADV_WILLNEED);
1573 case LINUX_POSIX_FADV_DONTNEED:
1574 return (POSIX_FADV_DONTNEED);
1575 case LINUX_POSIX_FADV_NOREUSE:
1576 return (POSIX_FADV_NOREUSE);
1583 linux_fadvise64(struct thread *td, struct linux_fadvise64_args *args)
1588 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1589 offset = PAIR32TO64(off_t, args->offset);
1591 offset = args->offset;
1594 advice = convert_fadvice(args->advice);
1597 return (kern_posix_fadvise(td, args->fd, offset, args->len, advice));
1600 #if defined(__i386__) || (defined(__amd64__) && defined(COMPAT_LINUX32))
1602 linux_fadvise64_64(struct thread *td, struct linux_fadvise64_64_args *args)
1607 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1608 len = PAIR32TO64(off_t, args->len);
1609 offset = PAIR32TO64(off_t, args->offset);
1612 offset = args->offset;
1615 advice = convert_fadvice(args->advice);
1618 return (kern_posix_fadvise(td, args->fd, offset, len, advice));
1620 #endif /* __i386__ || (__amd64__ && COMPAT_LINUX32) */
1622 #ifdef LINUX_LEGACY_SYSCALLS
1624 linux_pipe(struct thread *td, struct linux_pipe_args *args)
1629 error = kern_pipe(td, fildes, 0, NULL, NULL);
1633 error = copyout(fildes, args->pipefds, sizeof(fildes));
1635 (void)kern_close(td, fildes[0]);
1636 (void)kern_close(td, fildes[1]);
1644 linux_pipe2(struct thread *td, struct linux_pipe2_args *args)
1649 if ((args->flags & ~(LINUX_O_NONBLOCK | LINUX_O_CLOEXEC)) != 0)
1653 if ((args->flags & LINUX_O_NONBLOCK) != 0)
1654 flags |= O_NONBLOCK;
1655 if ((args->flags & LINUX_O_CLOEXEC) != 0)
1657 error = kern_pipe(td, fildes, flags, NULL, NULL);
1661 error = copyout(fildes, args->pipefds, sizeof(fildes));
1663 (void)kern_close(td, fildes[0]);
1664 (void)kern_close(td, fildes[1]);
1671 linux_dup3(struct thread *td, struct linux_dup3_args *args)
1676 if (args->oldfd == args->newfd)
1678 if ((args->flags & ~LINUX_O_CLOEXEC) != 0)
1680 if (args->flags & LINUX_O_CLOEXEC)
1681 cmd = F_DUP2FD_CLOEXEC;
1685 newfd = args->newfd;
1686 return (kern_fcntl(td, args->oldfd, cmd, newfd));
1690 linux_fallocate(struct thread *td, struct linux_fallocate_args *args)
1695 * We emulate only posix_fallocate system call for which
1698 if (args->mode != 0)
1699 return (EOPNOTSUPP);
1701 #if defined(__amd64__) && defined(COMPAT_LINUX32)
1702 len = PAIR32TO64(off_t, args->len);
1703 offset = PAIR32TO64(off_t, args->offset);
1706 offset = args->offset;
1709 return (kern_posix_fallocate(td, args->fd, offset, len));
1713 linux_copy_file_range(struct thread *td, struct linux_copy_file_range_args
1716 l_loff_t inoff, outoff, *inoffp, *outoffp;
1720 * copy_file_range(2) on Linux doesn't define any flags (yet), so is
1721 * the native implementation. Enforce it.
1723 if (args->flags != 0) {
1724 linux_msg(td, "copy_file_range unsupported flags 0x%x",
1729 inoffp = outoffp = NULL;
1730 if (args->off_in != NULL) {
1731 error = copyin(args->off_in, &inoff, sizeof(l_loff_t));
1736 if (args->off_out != NULL) {
1737 error = copyin(args->off_out, &outoff, sizeof(l_loff_t));
1743 error = kern_copy_file_range(td, args->fd_in, inoffp, args->fd_out,
1744 outoffp, args->len, flags);
1745 if (error == 0 && args->off_in != NULL)
1746 error = copyout(inoffp, args->off_in, sizeof(l_loff_t));
1747 if (error == 0 && args->off_out != NULL)
1748 error = copyout(outoffp, args->off_out, sizeof(l_loff_t));
1752 #define LINUX_MEMFD_PREFIX "memfd:"
1755 linux_memfd_create(struct thread *td, struct linux_memfd_create_args *args)
1757 char memfd_name[LINUX_NAME_MAX + 1];
1758 int error, flags, shmflags, oflags;
1761 * This is our clever trick to avoid the heap allocation to copy in the
1762 * uname. We don't really need to go this far out of our way, but it
1763 * does keep the rest of this function fairly clean as they don't have
1764 * to worry about cleanup on the way out.
1766 error = copyinstr(args->uname_ptr,
1767 memfd_name + sizeof(LINUX_MEMFD_PREFIX) - 1,
1768 LINUX_NAME_MAX - sizeof(LINUX_MEMFD_PREFIX) - 1, NULL);
1770 if (error == ENAMETOOLONG)
1775 memcpy(memfd_name, LINUX_MEMFD_PREFIX, sizeof(LINUX_MEMFD_PREFIX) - 1);
1776 flags = linux_to_bsd_bits(args->flags, mfd_bitmap, 0);
1777 if ((flags & ~(MFD_CLOEXEC | MFD_ALLOW_SEALING | MFD_HUGETLB |
1778 MFD_HUGE_MASK)) != 0)
1780 /* Size specified but no HUGETLB. */
1781 if ((flags & MFD_HUGE_MASK) != 0 && (flags & MFD_HUGETLB) == 0)
1783 /* We don't actually support HUGETLB. */
1784 if ((flags & MFD_HUGETLB) != 0)
1787 shmflags = SHM_GROW_ON_WRITE;
1788 if ((flags & MFD_CLOEXEC) != 0)
1789 oflags |= O_CLOEXEC;
1790 if ((flags & MFD_ALLOW_SEALING) != 0)
1791 shmflags |= SHM_ALLOW_SEALING;
1792 return (kern_shm_open2(td, SHM_ANON, oflags, 0, shmflags, NULL,
1797 linux_splice(struct thread *td, struct linux_splice_args *args)
1800 linux_msg(td, "syscall splice not really implemented");
1803 * splice(2) is documented to return EINVAL in various circumstances;
1804 * returning it instead of ENOSYS should hint the caller to use fallback
1811 linux_close_range(struct thread *td, struct linux_close_range_args *args)
1816 * Implementing close_range(CLOSE_RANGE_UNSHARE) allows Linux to
1817 * unshare filedesc table of the calling thread from others threads
1818 * in a thread group (i.e., process in the FreeBSD) or others processes,
1819 * which shares the same table, before closing the files. FreeBSD does
1820 * not have compatible unsharing mechanism due to the fact that sharing
1821 * process resources, including filedesc table, is at thread level in the
1822 * Linux, while in the FreeBSD it is at the process level.
1823 * Return EINVAL for now if the CLOSE_RANGE_UNSHARE flag is specified
1824 * until this new Linux API stabilizes.
1827 if ((args->flags & ~(LINUX_CLOSE_RANGE_CLOEXEC)) != 0)
1829 if (args->first > args->last)
1831 if ((args->flags & LINUX_CLOSE_RANGE_CLOEXEC) != 0)
1832 flags |= CLOSE_RANGE_CLOEXEC;
1833 return (kern_close_range(td, flags, args->first, args->last));