docs/TEMPORARY_DIRECTORIES.md

   1 ---
   2 title: Using /tmp/ And /var/tmp/ Safely
   3 ---
   4
   5 # Using `/tmp/` And `/var/tmp/` Safely
   6
   7 `/tmp/` and `/var/tmp/` are two world-writable directories Linux systems
   8 provide for temporary files. The former is typically on `tmpfs` and thus
   9 backed by RAM/swap, and flushed out on each reboot. The latter is typically a
  10 proper, persistent file system, and thus backed by physical storage. This
  11 means:
  12
  13 1. `/tmp/` should be used for smaller, size-bounded files only; `/var/tmp/`
  14    should be used for everything else.
  15
  16 2. Data that shall survive a boot cycle shouldn't be placed in `/tmp/`.
  17
  18 If the `$TMPDIR` environment variable is set, use that path, and neither use
  19 `/tmp/` nor `/var/tmp/` directly.
  20
  21 See
  22 [file-hierarchy(7)](https://www.freedesktop.org/software/systemd/man/file-hierarchy.html)
  23 for details about these two (and most other) directories of a Linux system.
  24
  25 ## Common Namespace
  26
  27 Note that `/tmp/` and `/var/tmp/` each define a common namespace shared by all
  28 local software. This means guessable file or directory names below either
  29 directory directly translate into a 🚨 Denial-of-Service (DoS) 🚨 vulnerability
  30 or worse: if some software creates a file or directory `/tmp/foo` then any
  31 other software that wants to create the same file or directory `/tmp/foo`
  32 either will fail (as the file already exists) or might be tricked into using
  33 untrusted files. Hence: do not use guessable names in `/tmp/` or `/var/tmp/` —
  34 if you do you open yourself up to a local DoS exploit or worse. (You can get
  35 away with using guessable names, if you pre-create subdirectories below `/tmp/`
  36 for them, like X11 does with `/tmp/.X11-unix/` through `tmpfiles.d/`
  37 drop-ins. However this is not recommended, as it is fully safe only if these
  38 directories are pre-created during early boot, and thus problematic if package
  39 installation during runtime is permitted.)
  40
  41 To protect yourself against these kinds of attacks Linux provides a couple of
  42 APIs that help you avoiding guessable names. Specifically:
  43
  44 1. Use [`mkstemp()`](http://man7.org/linux/man-pages/man3/mkstemp.3.html)
  45    (POSIX), `mkostemp()` (glibc),
  46    [`mkdtemp()`](http://man7.org/linux/man-pages/man3/mkdtemp.3.html) (POSIX),
  47    [`tmpfile()`](http://man7.org/linux/man-pages/man3/tmpfile.3.html) (C89)
  48
  49 2. Use [`open()`](http://man7.org/linux/man-pages/man2/open.2.html) with
  50    `O_TMPFILE` (Linux)
  51
  52 3. [`memfd_create()`](http://man7.org/linux/man-pages/man2/memfd_create.2.html)
  53    (Linux; this doesn't bother with `/tmp/` or `/var/tmp/` at all, but uses the
  54    same RAM/swap backing as `tmpfs` uses, hence is very similar to `/tmp/`
  55    semantics.)
  56
  57 For system services systemd provides the `PrivateTmp=` boolean setting. If
  58 turned on for a service (👍 which is highly recommended), `/tmp/` and
  59 `/var/tmp/` are replaced by private sub-directories, implemented through Linux
  60 file system namespacing and bind mounts. This means from the service's point of
  61 view `/tmp/` and `/var/tmp/` look and behave like they normally do, but in
  62 reality they are private sub-directories of the host's real `/tmp/` and
  63 `/var/tmp/`, and thus not system-wide locations anymore, but service-specific
  64 ones. This reduces the surface for local DoS attacks substantially. While it is
  65 recommended to turn this option on, it's highly recommended for applications
  66 not to rely on this solely to avoid DoS vulnerabilities, because this option is
  67 not available in environments where file system namespaces are prohibited, for
  68 example in certain container environments. This option is hence an extra line
  69 of defense, but should not be used as an excuse to rely on guessable names in
  70 `/tmp/` and `/var/tmp/`. When this option is used, the per-service temporary
  71 directories are removed whenever the service shuts down, hence the lifecycle of
  72 temporary files stored in it is substantially different from the case where
  73 this option is not used. Also note that some applications use `/tmp/` and
  74 `/var/tmp/` for sharing files and directories. If this option is turned on this
  75 is not possible anymore as after all each service gets its own instances of
  76 both directories.
  77
  78 ## Automatic Clean-Up
  79
  80 By default, `systemd-tmpfiles` will apply a concept of ⚠️ "ageing" to all files
  81 and directories stored in `/tmp/` and `/var/tmp/`. This means that files that
  82 have neither been changed nor read within a specific time frame are
  83 automatically removed in regular intervals. (This concept is not new to
  84 `systemd-tmpfiles` btw, it's inherited from previous subsystems such as
  85 `tmpwatch`.) By default files in `/tmp/` are cleaned up after 10 days, and
  86 those in `/var/tmp` after 30 days.
  87
  88 This automatic clean-up is important to ensure disk usage of these temporary
  89 directories doesn't grow without bounds, even when programs abort unexpectedly
  90 or otherwise don't clean up the temporary files/directories they create. On the
  91 other hand it creates problems for long-running software that does not expect
  92 temporary files it operates on to be suddenly removed. There are a couple of
  93 strategies to avoid these issues:
  94
  95 1. Make sure to always keep a file descriptor to the temporary files you
  96    operate on open, and only access the files through them. This way it doesn't
  97    matter whether the files have been unlinked from the file system: as long as
  98    you have the file descriptor open you can still access the file for both
  99    reading and writing. When operating this way it is recommended to delete the
 100    files right after creating them to ensure that on unexpected program
 101    termination the files or directories are implicitly released by the kernel.
 102
 103 2. 🥇 Use `memfd_create()` or `O_TMPFILE`. This is an extension of the
 104    suggestion above: files created this way are never linked under a filename
 105    in the file system. This means they are not subject to ageing (as they come
 106    unlinked out of the box), and there's no time window where a directory entry
 107    for the file exists in the file system, and thus behaviour is fully robust
 108    towards unexpected program termination as there are never files on disk that
 109    need to be explicitly deleted.
 110
 111 3. 🥇 Operate below a sub-directory of `/tmp/` and `/var/tmp/` you created, and
 112    take a BSD file lock ([`flock(dir_fd,
 113    LOCK_SH)`](http://man7.org/linux/man-pages/man2/flock.2.html)) on that
 114    sub-directory. This is particularly interesting when operating on more than
 115    a single file, or on file nodes that are not plain regular files, for
 116    example when extracting a tarball to a temporary directory. The ageing
 117    algorithm will skip all directories (and everything below them) that are
 118    locked through a BSD file lock. As BSD file locks are automatically released
 119    when the file descriptor they are taken on is closed, and all file
 120    descriptors opened by a process are implicitly closed when it exits, this is
 121    a robust mechanism that ensures all temporary files are subject to ageing
 122    when the program that owns them dies, but not while it is still running. Use
 123    this when decompressing tarballs that contain files with old
 124    modification/access times, as extracted files are otherwise immediately
 125    candidates for deletion by the ageing algorithm. The
 126    [`flock`](http://man7.org/linux/man-pages/man1/flock.1.html) tool of the
 127    `util-linux` packages makes this concept available to shell scripts. Note
 128    that `systemd-tmpfiles` only checks for BSD file locks on directories, locks
 129    on other types of file nodes (including regular files) are not considered.
 130
 131 4. Keep the access time of all temporary files created current. In regular
 132    intervals, use `utimensat()` or a related call to update the access time
 133    ("atime") of all files that shall be kept around. Since the ageing algorithm
 134    looks at the access time of files when deciding whether to delete them, it's
 135    sufficient to update their access times in sufficiently frequent intervals to
 136    ensure the files are not deleted. Since most applications (and tools such as
 137    `ls`) primarily care for the modification time (rather than the access time)
 138    using the access time for this purpose should be acceptable.
 139
 140 5. Set the "sticky" bit on regular files. The ageing logic skips deletion of
 141    all regular files that have the sticky bit (`chmod +t`) set. This is
 142    honoured for regular files only however, and has no effect on directories as
 143    the sticky bit has a different meaning for them.
 144
 145 6. Don't use `/tmp/` or `/var/tmp/`, but use your own sub-directory under
 146    `/run/` or `$XDG_RUNTIME_DIRECTORY` (the former if privileged, the latter if
 147    unprivileged), or `/var/lib/` and `~/.config/` (similar, but with
 148    persistency and suitable for larger data). The two temporary directories
 149    `/tmp/` and `/var/tmp/` come with the implicit clean-up semantics described
 150    above. When this is not desired, it's possible to create private per-package
 151    runtime or state directories, and place all temporary files there. However,
 152    do note that this means opting out of any kind of automatic clean-up, and it
 153    is hence particularly essential that the program cleans up generated files
 154    in these directories when they are no longer needed, in particular when the
 155    program dies unexpectedly. Note: this strategy is only really suitable for
 156    packages that operate in a "system wide singleton" fashion with "long"
 157    persistance of its data or state, i.e. as opposed to programs that run in
 158    multiple parallel or short-living instances. This is because a private
 159    directory under `/run` (and the other mentioned directories) is itself
 160    system and package specific singleton with greater longevity.
 161
 162 5. Exclude your temporary files from clean-ups via a `tmpfiles.d/` drop-in
 163    (which includes drop-ins in the runtime-only directory
 164    `/run/tmpfiles.d/`). The `x`/`X` line types may be used to exclude files
 165    matching the specified globbing patterns from the ageing logic. If this is
 166    used, automatic clean-up is not done for matching files and directory, and
 167    much like with the previous option it's hence essential that the program
 168    generating these temporary files carefully removes the temporary files it
 169    creates again, and in particular so if it dies unexpectedly.
 170
 171 🥇 The semantics of options 2 (in case you only deal with temporary files, not
 172 directories) and 3 (in case you deal with both) in the list above are in most
 173 cases the most preferable. It is thus recommended to stick to these two
 174 options.
 175
 176 While the ageing logic is very useful as a safety concept to ensure unused
 177 files and directories are eventually removed a well written program avoids even
 178 creating files that need such a clean-up. In particular:
 179
 180 1. Use `memfd_create()` or `O_TMPFILE` when creating temporary files.
 181
 182 2. `unlink()` temporary files right after creating them. This is very similar
 183    to `O_TMPFILE` behaviour: consider deleting temporary files right after
 184    creating them, while keeping open a file descriptor to them. Unlike
 185    `O_TMPFILE` this method also works on older Linux systems and other OSes
 186    that do not implement `O_TMPFILE`.
 187
 188 ## Disk Quota
 189
 190 Generally, files allocated from `/tmp/` and `/var/tmp/` are allocated from a
 191 pool shared by all local users. Moreover the space available in `/tmp/` is
 192 generally more restricted than `/var/tmp/`. This means, that in particular in
 193 `/tmp/` space should be considered scarce, and programs need to be prepared
 194 that no space is available. Essential programs might require a fallback logic
 195 using a different location for storing temporary files hence. Non-essential
 196 programs at least need to be prepared for `ENOSPC` errors and generate useful,
 197 actionable error messages.
 198
 199 Some setups employ per-user quota on `/var/tmp/` and possibly `/tmp/`, to make
 200 `ENOSPC` situations less likely, and harder to trigger from unprivileged
 201 users. However, in the general case no such per-user quota is implemented
 202 though, in particular not when `tmpfs` is used as backing file system, because
 203 — even today — `tmpfs` still provides no native quota support in the kernel.
 204
 205 ## Early Boot Considerations
 206
 207 Both `/tmp/` and `/var/tmp/` are not necessarily available during early boot,
 208 or — if they are available early — are not writable. This means software that
 209 is intended to run during early boot (i.e. before `basic.target` — or more
 210 specifically `local-fs.target` — is up) should not attempt to make use of
 211 either. Interfaces such as `memfd_create()` or files below a package-specific
 212 directory in `/run/` are much better options in this case. (Note that some
 213 packages instead use `/dev/shm/` for temporary files during early boot; this is
 214 not advisable however, as it offers no benefits over a private directory in
 215 `/run/` as both are backed by the same concept: `tmpfs`. The directory
 216 `/dev/shm/` exists to back POSIX shared memory (see
 217 [`shm_open()`](http://man7.org/linux/man-pages/man3/shm_open.3.html) and
 218 related calls), and not as a place for temporary files. `/dev/shm` is
 219 problematic as it is world-writable and there's no automatic clean-up logic in
 220 place.)