man7/path_resolution.7

   1 .\" Copyright (C) 2003 Andries Brouwer (aeb@cwi.nl)
   2 .\"
   3 .\" %%%LICENSE_START(VERBATIM)
   4 .\" Permission is granted to make and distribute verbatim copies of this
   5 .\" manual provided the copyright notice and this permission notice are
   6 .\" preserved on all copies.
   7 .\"
   8 .\" Permission is granted to copy and distribute modified versions of this
   9 .\" manual under the conditions for verbatim copying, provided that the
  10 .\" entire resulting derived work is distributed under the terms of a
  11 .\" permission notice identical to this one.
  12 .\"
  13 .\" Since the Linux kernel and libraries are constantly changing, this
  14 .\" manual page may be incorrect or out-of-date.  The author(s) assume no
  15 .\" responsibility for errors or omissions, or for damages resulting from
  16 .\" the use of the information contained herein.  The author(s) may not
  17 .\" have taken the same level of care in the production of this manual,
  18 .\" which is licensed free of charge, as they might when working
  19 .\" professionally.
  20 .\"
  21 .\" Formatted or processed versions of this manual, if unaccompanied by
  22 .\" the source, must acknowledge the copyright and authors of this work.
  23 .\" %%%LICENSE_END
  24 .\"
  25 .TH PATH_RESOLUTION 7 2017-11-26 "Linux" "Linux Programmer's Manual"
  26 .SH NAME
  27 path_resolution \- how a pathname is resolved to a file
  28 .SH DESCRIPTION
  29 Some UNIX/Linux system calls have as parameter one or more filenames.
  30 A filename (or pathname) is resolved as follows.
  31 .SS Step 1: start of the resolution process
  32 If the pathname starts with the \(aq/\(aq character,
  33 the starting lookup directory
  34 is the root directory of the calling process.
  35 (A process inherits its
  36 root directory from its parent.
  37 Usually this will be the root directory
  38 of the file hierarchy.
  39 A process may get a different root directory
  40 by use of the
  41 .BR chroot (2)
  42 system call.
  43 A process may get an entirely private mount namespace in case
  44 it\(emor one of its ancestors\(emwas started by an invocation of the
  45 .BR clone (2)
  46 system call that had the
  47 .B CLONE_NEWNS
  48 flag set.)
  49 This handles the \(aq/\(aq part of the pathname.
  50 .PP
  51 If the pathname does not start with the \(aq/\(aq character, the
  52 starting lookup directory of the resolution process is the current working
  53 directory of the process.
  54 (This is also inherited from the parent.
  55 It can be changed by use of the
  56 .BR chdir (2)
  57 system call.)
  58 .PP
  59 Pathnames starting with a \(aq/\(aq character are called absolute pathnames.
  60 Pathnames not starting with a \(aq/\(aq are called relative pathnames.
  61 .SS Step 2: walk along the path
  62 Set the current lookup directory to the starting lookup directory.
  63 Now, for each nonfinal component of the pathname, where a component
  64 is a substring delimited by \(aq/\(aq characters, this component is looked up
  65 in the current lookup directory.
  66 .PP
  67 If the process does not have search permission on
  68 the current lookup directory,
  69 an
  70 .B EACCES
  71 error is returned ("Permission denied").
  72 .PP
  73 If the component is not found, an
  74 .B ENOENT
  75 error is returned
  76 ("No such file or directory").
  77 .PP
  78 If the component is found, but is neither a directory nor a symbolic link,
  79 an
  80 .B ENOTDIR
  81 error is returned ("Not a directory").
  82 .PP
  83 If the component is found and is a directory, we set the
  84 current lookup directory to that directory, and go to the
  85 next component.
  86 .PP
  87 If the component is found and is a symbolic link (symlink), we first
  88 resolve this symbolic link (with the current lookup directory
  89 as starting lookup directory).
  90 Upon error, that error is returned.
  91 If the result is not a directory, an
  92 .B ENOTDIR
  93 error is returned.
  94 If the resolution of the symlink is successful and returns a directory,
  95 we set the current lookup directory to that directory, and go to
  96 the next component.
  97 Note that the resolution process here can involve recursion if the
  98 prefix ('dirname') component of a pathname contains a filename
  99 that is a symbolic link that resolves to a directory (where the
 100 prefix component of that directory may contain a symbolic link, and so on).
 101 In order to protect the kernel against stack overflow, and also
 102 to protect against denial of service, there are limits on the
 103 maximum recursion depth, and on the maximum number of symbolic links
 104 followed.
 105 An
 106 .B ELOOP
 107 error is returned when the maximum is
 108 exceeded ("Too many levels of symbolic links").
 109 .PP
 110 .\"
 111 .\" presently: max recursion depth during symlink resolution: 5
 112 .\" max total number of symbolic links followed: 40
 113 .\" _POSIX_SYMLOOP_MAX is 8
 114 As currently implemented on Linux, the maximum number
 115 .\" MAXSYMLINKS is 40
 116 of symbolic links that will be followed while resolving a pathname is 40.
 117 In kernels before 2.6.18, the limit on the recursion depth was 5.
 118 Starting with Linux 2.6.18, this limit
 119 .\" MAX_NESTED_LINKS
 120 was raised to 8.
 121 In Linux 4.2,
 122 .\" commit 894bc8c4662ba9daceafe943a5ba0dd407da5cd3
 123 the kernel's pathname-resolution code
 124 was reworked to eliminate the use of recursion,
 125 so that the only limit that remains is the maximum of 40
 126 resolutions for the entire pathname.
 127 .SS Step 3: find the final entry
 128 The lookup of the final component of the pathname goes just like
 129 that of all other components, as described in the previous step,
 130 with two differences: (i) the final component need not be a
 131 directory (at least as far as the path resolution process is
 132 concerned\(emit may have to be a directory, or a nondirectory, because of
 133 the requirements of the specific system call), and (ii) it
 134 is not necessarily an error if the component is not found\(emmaybe
 135 we are just creating it.
 136 The details on the treatment
 137 of the final entry are described in the manual pages of the specific
 138 system calls.
 139 .SS . and ..
 140 By convention, every directory has the entries "." and "..",
 141 which refer to the directory itself and to its parent directory,
 142 respectively.
 143 .PP
 144 The path resolution process will assume that these entries have
 145 their conventional meanings, regardless of whether they are
 146 actually present in the physical filesystem.
 147 .PP
 148 One cannot walk down past the root: "/.." is the same as "/".
 149 .SS Mount points
 150 After a "mount dev path" command, the pathname "path" refers to
 151 the root of the filesystem hierarchy on the device "dev", and no
 152 longer to whatever it referred to earlier.
 153 .PP
 154 One can walk out of a mounted filesystem: "path/.." refers to
 155 the parent directory of "path",
 156 outside of the filesystem hierarchy on "dev".
 157 .SS Trailing slashes
 158 If a pathname ends in a \(aq/\(aq, that forces resolution of the preceding
 159 component as in Step 2: it has to exist and resolve to a directory.
 160 Otherwise, a trailing \(aq/\(aq is ignored.
 161 (Or, equivalently, a pathname with a trailing \(aq/\(aq is equivalent to
 162 the pathname obtained by appending \(aq.\(aq to it.)
 163 .SS Final symlink
 164 If the last component of a pathname is a symbolic link, then it
 165 depends on the system call whether the file referred to will be
 166 the symbolic link or the result of path resolution on its contents.
 167 For example, the system call
 168 .BR lstat (2)
 169 will operate on the symlink, while
 170 .BR stat (2)
 171 operates on the file pointed to by the symlink.
 172 .SS Length limit
 173 There is a maximum length for pathnames.
 174 If the pathname (or some
 175 intermediate pathname obtained while resolving symbolic links)
 176 is too long, an
 177 .B ENAMETOOLONG
 178 error is returned ("Filename too long").
 179 .SS Empty pathname
 180 In the original UNIX, the empty pathname referred to the current directory.
 181 Nowadays POSIX decrees that an empty pathname must not be resolved
 182 successfully.
 183 Linux returns
 184 .B ENOENT
 185 in this case.
 186 .SS Permissions
 187 The permission bits of a file consist of three groups of three bits; see
 188 .BR chmod (1)
 189 and
 190 .BR stat (2).
 191 The first group of three is used when the effective user ID of
 192 the calling process equals the owner ID of the file.
 193 The second group
 194 of three is used when the group ID of the file either equals the
 195 effective group ID of the calling process, or is one of the
 196 supplementary group IDs of the calling process (as set by
 197 .BR setgroups (2)).
 198 When neither holds, the third group is used.
 199 .PP
 200 Of the three bits used, the first bit determines read permission,
 201 the second write permission, and the last execute permission
 202 in case of ordinary files, or search permission in case of directories.
 203 .PP
 204 Linux uses the fsuid instead of the effective user ID in permission checks.
 205 Ordinarily the fsuid will equal the effective user ID, but the fsuid can be
 206 changed by the system call
 207 .BR setfsuid (2).
 208 .PP
 209 (Here "fsuid" stands for something like "filesystem user ID".
 210 The concept was required for the implementation of a user space
 211 NFS server at a time when processes could send a signal to a process
 212 with the same effective user ID.
 213 It is obsolete now.
 214 Nobody should use
 215 .BR setfsuid (2).)
 216 .PP
 217 Similarly, Linux uses the fsgid ("filesystem group ID")
 218 instead of the effective group ID.
 219 See
 220 .BR setfsgid (2).
 221 .\" FIXME . say something about filesystem mounted read-only ?
 222 .SS Bypassing permission checks: superuser and capabilities
 223 On a traditional UNIX system, the superuser
 224 .RI ( root ,
 225 user ID 0) is all-powerful, and bypasses all permissions restrictions
 226 when accessing files.
 227 .\" (but for exec at least one x bit must be set) -- AEB
 228 .\" but there is variation across systems on this point: for
 229 .\" example, HP-UX and Tru64 are as described by AEB.  However,
 230 .\" on some implementations (e.g., Solaris, FreeBSD),
 231 .\" access(X_OK) by superuser will report success, regardless
 232 .\" of the file's execute permission bits. -- MTK (Oct 05)
 233 .PP
 234 On Linux, superuser privileges are divided into capabilities (see
 235 .BR capabilities (7)).
 236 Two capabilities are relevant for file permissions checks:
 237 .B CAP_DAC_OVERRIDE
 238 and
 239 .BR CAP_DAC_READ_SEARCH .
 240 (A process has these capabilities if its fsuid is 0.)
 241 .PP
 242 The
 243 .B CAP_DAC_OVERRIDE
 244 capability overrides all permission checking,
 245 but grants execute permission only when at least one
 246 of the file's three execute permission bits is set.
 247 .PP
 248 The
 249 .B CAP_DAC_READ_SEARCH
 250 capability grants read and search permission
 251 on directories, and read permission on ordinary files.
 252 .\" FIXME . say something about immutable files
 253 .\" FIXME . say something about ACLs
 254 .SH SEE ALSO
 255 .BR readlink (2),
 256 .BR capabilities (7),
 257 .BR credentials (7),
 258 .BR symlink (7)