Some UNIX/Linux system calls have as parameter one or more
filenames. A filename (or pathname) is resolved as follows.
Step 1: start of the resolution process
If the pathname starts with the '/' character, the starting
lookup directory is the root directory of the calling process. A
process inherits its root directory from its parent. Usually
this will be the root directory of the file hierarchy. A process
may get a different root directory by use of the chroot(2) system
call, or may temporarily use a different root directory by using
openat2(2) with the RESOLVE_IN_ROOT flag set.
A process may get an entirely private mount namespace in case it—
or one of its ancestors—was started by an invocation of the
clone(2) system call that had the CLONE_NEWNS flag set. This
handles the '/' part of the pathname.
If the pathname does not start with the '/' character, the
starting lookup directory of the resolution process is the
current working directory of the process — or in the case of
openat(2)-style system calls, the dfd argument (or the current
working directory if AT_FDCWD is passed as the dfd argument).
The current working directory is inherited from the parent, and
can be changed by use of the chdir(2) system call.
Pathnames starting with a '/' character are called absolute
pathnames. Pathnames not starting with a '/' are called relative
pathnames.
Step 2: walk along the path
Set the current lookup directory to the starting lookup
directory. Now, for each nonfinal component of the pathname,
where a component is a substring delimited by '/' characters,
this component is looked up in the current lookup directory.
If the process does not have search permission on the current
lookup directory, an EACCES error is returned ("Permission
denied").
If the component is not found, an ENOENT error is returned ("No
such file or directory").
If the component is found, but is neither a directory nor a
symbolic link, an ENOTDIR error is returned ("Not a directory").
If the component is found and is a directory, we set the current
lookup directory to that directory, and go to the next component.
If the component is found and is a symbolic link (symlink), we
first resolve this symbolic link (with the current lookup
directory as starting lookup directory). Upon error, that error
is returned. If the result is not a directory, an ENOTDIR error
is returned. If the resolution of the symbolic link is
successful and returns a directory, we set the current lookup
directory to that directory, and go to the next component. Note
that the resolution process here can involve recursion if the
prefix ('dirname') component of a pathname contains a filename
that is a symbolic link that resolves to a directory (where the
prefix component of that directory may contain a symbolic link,
and so on). In order to protect the kernel against stack
overflow, and also to protect against denial of service, there
are limits on the maximum recursion depth, and on the maximum
number of symbolic links followed. An ELOOP error is returned
when the maximum is exceeded ("Too many levels of symbolic
links").
As currently implemented on Linux, the maximum number of symbolic
links that will be followed while resolving a pathname is 40. In
kernels before 2.6.18, the limit on the recursion depth was 5.
Starting with Linux 2.6.18, this limit was raised to 8. In Linux
4.2, the kernel's pathname-resolution code was reworked to
eliminate the use of recursion, so that the only limit that
remains is the maximum of 40 resolutions for the entire pathname.
The resolution of symbolic links during this stage can be blocked
by using openat2(2), with the RESOLVE_NO_SYMLINKS flag set.
Step 3: find the final entry
The lookup of the final component of the pathname goes just like
that of all other components, as described in the previous step,
with two differences: (i) the final component need not be a
directory (at least as far as the path resolution process is
concerned—it may have to be a directory, or a nondirectory,
because of the requirements of the specific system call), and
(ii) it is not necessarily an error if the component is not
found—maybe we are just creating it. The details on the
treatment of the final entry are described in the manual pages of
the specific system calls.
. and ..
By convention, every directory has the entries "." and "..",
which refer to the directory itself and to its parent directory,
respectively.
The path resolution process will assume that these entries have
their conventional meanings, regardless of whether they are
actually present in the physical filesystem.
One cannot walk up past the root: "/.." is the same as "/".
Mount points
After a "mount dev path" command, the pathname "path" refers to
the root of the filesystem hierarchy on the device "dev", and no
longer to whatever it referred to earlier.
One can walk out of a mounted filesystem: "path/.." refers to the
parent directory of "path", outside of the filesystem hierarchy
on "dev".
Traversal of mount points can be blocked by using openat2(2),
with the RESOLVE_NO_XDEV flag set (though note that this also
restricts bind mount traversal).
Trailing slashes
If a pathname ends in a '/', that forces resolution of the
preceding component as in Step 2: the component preceding the
slash either exists and resolves to a directory or it names a
directory that is to be created immediately after the pathname is
resolved. Otherwise, a trailing '/' is ignored.
Final symlink
If the last component of a pathname is a symbolic link, then it
depends on the system call whether the file referred to will be
the symbolic link or the result of path resolution on its
contents. For example, the system call lstat(2) will operate on
the symlink, while stat(2) operates on the file pointed to by the
symlink.
Length limit
There is a maximum length for pathnames. If the pathname (or
some intermediate pathname obtained while resolving symbolic
links) is too long, an ENAMETOOLONG error is returned ("Filename
too long").
Empty pathname
In the original UNIX, the empty pathname referred to the current
directory. Nowadays POSIX decrees that an empty pathname must
not be resolved successfully. Linux returns ENOENT in this case.
Permissions
The permission bits of a file consist of three groups of three
bits; see chmod(1) and stat(2). The first group of three is used
when the effective user ID of the calling process equals the
owner ID of the file. The second group of three is used when the
group ID of the file either equals the effective group ID of the
calling process, or is one of the supplementary group IDs of the
calling process (as set by setgroups(2)). When neither holds,
the third group is used.
Of the three bits used, the first bit determines read permission,
the second write permission, and the last execute permission in
case of ordinary files, or search permission in case of
directories.
Linux uses the fsuid instead of the effective user ID in
permission checks. Ordinarily the fsuid will equal the effective
user ID, but the fsuid can be changed by the system call
setfsuid(2).
(Here "fsuid" stands for something like "filesystem user ID".
The concept was required for the implementation of a user space
NFS server at a time when processes could send a signal to a
process with the same effective user ID. It is obsolete now.
Nobody should use setfsuid(2).)
Similarly, Linux uses the fsgid ("filesystem group ID") instead
of the effective group ID. See setfsgid(2).
Bypassing permission checks: superuser and capabilities
On a traditional UNIX system, the superuser (root, user ID 0) is
all-powerful, and bypasses all permissions restrictions when
accessing files.
On Linux, superuser privileges are divided into capabilities (see
capabilities(7)). Two capabilities are relevant for file
permissions checks: CAP_DAC_OVERRIDE and CAP_DAC_READ_SEARCH. (A
process has these capabilities if its fsuid is 0.)
The CAP_DAC_OVERRIDE capability overrides all permission
checking, but grants execute permission only when at least one of
the file's three execute permission bits is set.
The CAP_DAC_READ_SEARCH capability grants read and search
permission on directories, and read permission on ordinary files.
