Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

chown

CHOWN(2)                   Linux Programmer's Manual                  CHOWN(2)

NAME
       chown, fchown, lchown, fchownat - change ownership of a file

SYNOPSIS
       #include <unistd.h>

       int chown(const char *pathname, uid_t owner, gid_t group);
       int fchown(int fd, uid_t owner, gid_t group);
       int lchown(const char *pathname, uid_t owner, gid_t group);

       #include <fcntl.h>           /* Definition of AT_* constants */
       #include <unistd.h>

       int fchownat(int dirfd, const char *pathname,
                    uid_t owner, gid_t group, int flags);

   Feature Test Macro Requirements for glibc (see feature_test_macros(7)):

       fchown(), lchown():
           /* Since glibc 2.12: */ _POSIX_C_SOURCE >= 200809L
               || _XOPEN_SOURCE >= 500
               || /* Glibc versions <= 2.19: */ _BSD_SOURCE
       fchownat():
           Since glibc 2.10:
               _POSIX_C_SOURCE >= 200809L
           Before glibc 2.10:
               _ATFILE_SOURCE

DESCRIPTION
       These  system calls change the owner and group of a file.  The chown(),
       fchown(), and lchown() system calls differ only  in  how  the  file  is
       specified:

       * chown()  changes  the  ownership  of  the file specified by pathname,
         which is dereferenced if it is a symbolic link.

       * fchown() changes the ownership of the file referred to  by  the  open
         file descriptor fd.

       * lchown() is like chown(), but does not dereference symbolic links.

       Only  a  privileged  process (Linux: one with the CAP_CHOWN capability)
       may change the owner of a file.  The owner of a  file  may  change  the
       group  of  the  file  to  any group of which that owner is a member.  A
       privileged process (Linux: with CAP_CHOWN) may change the  group  arbi-
       trarily.

       If the owner or group is specified as -1, then that ID is not changed.

       When the owner or group of an executable file is changed by an unprivi-
       leged user, the S_ISUID and S_ISGID mode bits are cleared.  POSIX  does
       not specify whether this also should happen when root does the chown();
       the Linux behavior depends on  the  kernel  version,  and  since  Linux
       2.2.13,  root is treated like other users.  In case of a non-group-exe-
       cutable file (i.e., one for which the S_IXGRP bit is not set) the S_IS-
       GID bit indicates mandatory locking, and is not cleared by a chown().

       When the owner or group of an executable file is changed (by any user),
       all capability sets for the file are cleared.

   fchownat()
       The fchownat() system call operates in exactly the same way as chown(),
       except for the differences described here.

       If  the  pathname given in pathname is relative, then it is interpreted
       relative to the directory referred to  by  the  file  descriptor  dirfd
       (rather  than  relative to the current working directory of the calling
       process, as is done by chown() for a relative pathname).

       If pathname is relative and dirfd is the special value  AT_FDCWD,  then
       pathname  is  interpreted  relative to the current working directory of
       the calling process (like chown()).

       If pathname is absolute, then dirfd is ignored.

       The flags argument is a bit mask created by ORing together 0 or more of
       the following values;

       AT_EMPTY_PATH (since Linux 2.6.39)
              If  pathname is an empty string, operate on the file referred to
              by dirfd (which may have been obtained using the open(2)  O_PATH
              flag).   In  this case, dirfd can refer to any type of file, not
              just a directory.  If dirfd is AT_FDCWD, the  call  operates  on
              the current working directory.  This flag is Linux-specific; de-
              fine _GNU_SOURCE to obtain its definition.

       AT_SYMLINK_NOFOLLOW
              If pathname is a symbolic link, do not dereference  it:  instead
              operate  on the link itself, like lchown().  (By default, fchow-
              nat() dereferences symbolic links, like chown().)

       See openat(2) for an explanation of the need for fchownat().

RETURN VALUE
       On success, zero is returned.  On error, -1 is returned, and  errno  is
       set appropriately.

ERRORS
       Depending  on  the filesystem, errors other than those listed below can
       be returned.

       The more general errors for chown() are listed below.

       EACCES Search permission is denied on a component of the  path  prefix.
              (See also path_resolution(7).)

       EFAULT pathname points outside your accessible address space.

       ELOOP  Too many symbolic links were encountered in resolving pathname.

       ENAMETOOLONG
              pathname is too long.

       ENOENT The file does not exist.

       ENOMEM Insufficient kernel memory was available.

       ENOTDIR
              A component of the path prefix is not a directory.

       EPERM  The  calling  process did not have the required permissions (see
              above) to change owner and/or group.

       EPERM  The   file   is   marked   immutable   or   append-only.    (See
              ioctl_iflags(2).)

       EROFS  The named file resides on a read-only filesystem.

       The general errors for fchown() are listed below:

       EBADF  fd is not a valid open file descriptor.

       EIO    A low-level I/O error occurred while modifying the inode.

       ENOENT See above.

       EPERM  See above.

       EROFS  See above.

       The  same  errors that occur for chown() can also occur for fchownat().
       The following additional errors can occur for fchownat():

       EBADF  dirfd is not a valid file descriptor.

       EINVAL Invalid flag specified in flags.

       ENOTDIR
              pathname is relative and dirfd is a file descriptor referring to
              a file other than a directory.

VERSIONS
       fchownat()  was  added  to  Linux in kernel 2.6.16; library support was
       added to glibc in version 2.4.

CONFORMING TO
       chown(), fchown(), lchown(): 4.4BSD, SVr4, POSIX.1-2001, POSIX.1-2008.

       The 4.4BSD version can be used only by the superuser (that is, ordinary
       users cannot give away files).

       fchownat(): POSIX.1-2008.

NOTES
   Ownership of new files
       When  a new file is created (by, for example, open(2) or mkdir(2)), its
       owner is made the same as  the  filesystem  user  ID  of  the  creating
       process.   The group of the file depends on a range of factors, includ-
       ing the type of filesystem, the options used to mount  the  filesystem,
       and  whether  or not the set-group-ID mode bit is enabled on the parent
       directory.  If the filesystem supports the -o grpid  (or,  synonymously
       -o bsdgroups)  and -o nogrpid (or, synonymously -o sysvgroups) mount(8)
       options, then the rules are as follows:

       * If the filesystem is mounted with -o grpid, then the group of  a  new
         file is made the same as that of the parent directory.

       * If the filesystem is mounted with -o nogrpid and the set-group-ID bit
         is disabled on the parent directory, then the group of a new file  is
         made the same as the process's filesystem GID.

       * If the filesystem is mounted with -o nogrpid and the set-group-ID bit
         is enabled on the parent directory, then the group of a new  file  is
         made the same as that of the parent directory.

       As  at  Linux  4.12, the -o grpid and -o nogrpid mount options are sup-
       ported by ext2, ext3, ext4, and XFS.  Filesystems  that  don't  support
       these mount options follow the -o nogrpid rules.

   Glibc notes
       On  older  kernels  where  fchownat() is unavailable, the glibc wrapper
       function falls back to the use of chown() and lchown().  When  pathname
       is  a  relative pathname, glibc constructs a pathname based on the sym-
       bolic link in /proc/self/fd that corresponds to the dirfd argument.

   NFS
       The chown() semantics are  deliberately  violated  on  NFS  filesystems
       which  have  UID  mapping  enabled.  Additionally, the semantics of all
       system calls which access  the  file  contents  are  violated,  because
       chown()  may  cause  immediate access revocation on already open files.
       Client side caching may lead to a delay between the time  where  owner-
       ship  have  been  changed to allow access for a user and the time where
       the file can actually be accessed by the user on other clients.

   Historical details
       The original Linux chown(), fchown(), and lchown()  system  calls  sup-
       ported  only  16-bit user and group IDs.  Subsequently, Linux 2.4 added
       chown32(), fchown32(), and  lchown32(),  supporting  32-bit  IDs.   The
       glibc  chown(),  fchown(), and lchown() wrapper functions transparently
       deal with the variations across kernel versions.

       In versions of Linux  prior  to  2.1.81  (and  distinct  from  2.1.46),
       chown()  did  not  follow  symbolic links.  Since Linux 2.1.81, chown()
       does follow symbolic links, and there is a  new  system  call  lchown()
       that does not follow symbolic links.  Since Linux 2.1.86, this new call
       (that has the same semantics as the  old  chown())  has  got  the  same
       syscall number, and chown() got the newly introduced number.

EXAMPLE
       The  following  program  changes the ownership of the file named in its
       second command-line argument to the value specified in its  first  com-
       mand-line argument.  The new owner can be specified either as a numeric
       user ID, or as a username (which is converted to a  user  ID  by  using
       getpwnam(3) to perform a lookup in the system password file).

   Program source
       #include <pwd.h>
       #include <stdio.h>
       #include <stdlib.h>
       #include <unistd.h>

       int
       main(int argc, char *argv[])
       {
           uid_t uid;
           struct passwd *pwd;
           char *endptr;

           if (argc != 3 || argv[1][0] == '\0') {
               fprintf(stderr, "%s <owner> <file>\n", argv[0]);
               exit(EXIT_FAILURE);
           }

           uid = strtol(argv[1], &endptr, 10);  /* Allow a numeric string */

           if (*endptr != '\0') {         /* Was not pure numeric string */
               pwd = getpwnam(argv[1]);   /* Try getting UID for username */
               if (pwd == NULL) {
                   perror("getpwnam");
                   exit(EXIT_FAILURE);
               }

               uid = pwd->pw_uid;
           }

           if (chown(argv[2], uid, -1) == -1) {
               perror("chown");
               exit(EXIT_FAILURE);
           }

           exit(EXIT_SUCCESS);
       }

SEE ALSO
       chgrp(1), chown(1), chmod(2), flock(2), path_resolution(7), symlink(7)

COLOPHON
       This  page  is  part of release 5.05 of the Linux man-pages project.  A
       description of the project, information about reporting bugs,  and  the
       latest     version     of     this    page,    can    be    found    at
       https://www.kernel.org/doc/man-pages/.

Linux                             2019-03-06                          CHOWN(2)
Man Pages Copyright Respective Owners. Site Copyright (C) 1994 - 2024 Hurricane Electric. All Rights Reserved.