1PATH_RESOLUTION(7) Linux Programmer's Manual PATH_RESOLUTION(7)
2
6 path_resolution - how a pathname is resolved to a file
7
9 Some UNIX/Linux system calls have as parameter one or more filenames.
10 A filename (or pathname) is resolved as follows.
11 Step 1: start of the resolution process
13 If the pathname starts with the '/' character, the starting lookup di‐
14 rectory is the root directory of the calling process. A process inher‐
15 its its root directory from its parent. Usually this will be the root
16 directory of the file hierarchy. A process may get a different root
17 directory by use of the chroot(2) system call, or may temporarily use a
18 different root directory by using openat2(2) with the RESOLVE_IN_ROOT
19 flag set.
20 A process may get an entirely private mount namespace in case it—or one
22 of its ancestors—was started by an invocation of the clone(2) system
23 call that had the CLONE_NEWNS flag set. This handles the '/' part of
24 the pathname.
25 If the pathname does not start with the '/' character, the starting
27 lookup directory of the resolution process is the current working di‐
28 rectory of the process — or in the case of openat(2)-style system
29 calls, the dfd argument (or the current working directory if AT_FDCWD
30 is passed as the dfd argument). The current working directory is in‐
31 herited from the parent, and can be changed by use of the chdir(2) sys‐
32 tem call.)
33 Pathnames starting with a '/' character are called absolute pathnames.
35 Pathnames not starting with a '/' are called relative pathnames.
36 Step 2: walk along the path
38 Set the current lookup directory to the starting lookup directory.
39 Now, for each nonfinal component of the pathname, where a component is
40 a substring delimited by '/' characters, this component is looked up in
41 the current lookup directory.
42 If the process does not have search permission on the current lookup
44 directory, an EACCES error is returned ("Permission denied").
45 If the component is not found, an ENOENT error is returned ("No such
47 file or directory").
48 If the component is found, but is neither a directory nor a symbolic
50 link, an ENOTDIR error is returned ("Not a directory").
51 If the component is found and is a directory, we set the current lookup
53 directory to that directory, and go to the next component.
54 If the component is found and is a symbolic link (symlink), we first
56 resolve this symbolic link (with the current lookup directory as start‐
57 ing lookup directory). Upon error, that error is returned. If the re‐
58 sult is not a directory, an ENOTDIR error is returned. If the resolu‐
59 tion of the symbolic link is successful and returns a directory, we set
60 the current lookup directory to that directory, and go to the next com‐
61 ponent. Note that the resolution process here can involve recursion if
62 the prefix ('dirname') component of a pathname contains a filename that
63 is a symbolic link that resolves to a directory (where the prefix com‐
64 ponent of that directory may contain a symbolic link, and so on). In
65 order to protect the kernel against stack overflow, and also to protect
66 against denial of service, there are limits on the maximum recursion
67 depth, and on the maximum number of symbolic links followed. An ELOOP
68 error is returned when the maximum is exceeded ("Too many levels of
69 symbolic links").
70 As currently implemented on Linux, the maximum number of symbolic links
72 that will be followed while resolving a pathname is 40. In kernels be‐
73 fore 2.6.18, the limit on the recursion depth was 5. Starting with
74 Linux 2.6.18, this limit was raised to 8. In Linux 4.2, the kernel's
75 pathname-resolution code was reworked to eliminate the use of recur‐
76 sion, so that the only limit that remains is the maximum of 40 resolu‐
77 tions for the entire pathname.
78 The resolution of symbolic links during this stage can be blocked by
80 using openat2(2), with the RESOLVE_NO_SYMLINKS flag set.
81 Step 3: find the final entry
83 The lookup of the final component of the pathname goes just like that
84 of all other components, as described in the previous step, with two
85 differences: (i) the final component need not be a directory (at least
86 as far as the path resolution process is concerned—it may have to be a
87 directory, or a nondirectory, because of the requirements of the spe‐
88 cific system call), and (ii) it is not necessarily an error if the com‐
89 ponent is not found—maybe we are just creating it. The details on the
90 treatment of the final entry are described in the manual pages of the
91 specific system calls.
92 . and ..
94 By convention, every directory has the entries "." and "..", which re‐
95 fer to the directory itself and to its parent directory, respectively.
96 The path resolution process will assume that these entries have their
98 conventional meanings, regardless of whether they are actually present
99 in the physical filesystem.
100 One cannot walk up past the root: "/.." is the same as "/".
102 Mount points
104 After a "mount dev path" command, the pathname "path" refers to the
105 root of the filesystem hierarchy on the device "dev", and no longer to
106 whatever it referred to earlier.
107 One can walk out of a mounted filesystem: "path/.." refers to the par‐
109 ent directory of "path", outside of the filesystem hierarchy on "dev".
110 Traversal of mount points can be blocked by using openat2(2), with the
112 RESOLVE_NO_XDEV flag set (though note that this also restricts bind
113 mount traversal).
114 Trailing slashes
116 If a pathname ends in a '/', that forces resolution of the preceding
117 component as in Step 2: it has to exist and resolve to a directory.
118 Otherwise, a trailing '/' is ignored. (Or, equivalently, a pathname
119 with a trailing '/' is equivalent to the pathname obtained by appending
120 '.' to it.)
121 Final symlink
123 If the last component of a pathname is a symbolic link, then it depends
124 on the system call whether the file referred to will be the symbolic
125 link or the result of path resolution on its contents. For example,
126 the system call lstat(2) will operate on the symlink, while stat(2) op‐
127 erates on the file pointed to by the symlink.
128 Length limit
130 There is a maximum length for pathnames. If the pathname (or some in‐
131 termediate pathname obtained while resolving symbolic links) is too
132 long, an ENAMETOOLONG error is returned ("Filename too long").
133 Empty pathname
135 In the original UNIX, the empty pathname referred to the current direc‐
136 tory. Nowadays POSIX decrees that an empty pathname must not be re‐
137 solved successfully. Linux returns ENOENT in this case.
138 Permissions
140 The permission bits of a file consist of three groups of three bits;
141 see chmod(1) and stat(2). The first group of three is used when the
142 effective user ID of the calling process equals the owner ID of the
143 file. The second group of three is used when the group ID of the file
144 either equals the effective group ID of the calling process, or is one
145 of the supplementary group IDs of the calling process (as set by set‐
146 groups(2)). When neither holds, the third group is used.
147 Of the three bits used, the first bit determines read permission, the
149 second write permission, and the last execute permission in case of or‐
150 dinary files, or search permission in case of directories.
151 Linux uses the fsuid instead of the effective user ID in permission
153 checks. Ordinarily the fsuid will equal the effective user ID, but the
154 fsuid can be changed by the system call setfsuid(2).
155 (Here "fsuid" stands for something like "filesystem user ID". The con‐
157 cept was required for the implementation of a user space NFS server at
158 a time when processes could send a signal to a process with the same
159 effective user ID. It is obsolete now. Nobody should use setf‐
160 suid(2).)
161 Similarly, Linux uses the fsgid ("filesystem group ID") instead of the
163 effective group ID. See setfsgid(2).
164 Bypassing permission checks: superuser and capabilities
166 On a traditional UNIX system, the superuser (root, user ID 0) is all-
167 powerful, and bypasses all permissions restrictions when accessing
168 files.
169 On Linux, superuser privileges are divided into capabilities (see capa‐
171 bilities(7)). Two capabilities are relevant for file permissions
172 checks: CAP_DAC_OVERRIDE and CAP_DAC_READ_SEARCH. (A process has these
173 capabilities if its fsuid is 0.)
174 The CAP_DAC_OVERRIDE capability overrides all permission checking, but
176 grants execute permission only when at least one of the file's three
177 execute permission bits is set.
178 The CAP_DAC_READ_SEARCH capability grants read and search permission on
180 directories, and read permission on ordinary files.
181
183 readlink(2), capabilities(7), credentials(7), symlink(7)
184
186 This page is part of release 5.12 of the Linux man-pages project. A
187 description of the project, information about reporting bugs, and the
188 latest version of this page, can be found at
189 https://www.kernel.org/doc/man-pages/.
190Linux 2020-04-11 PATH_RESOLUTION(7)