docs/TEMPORARY_DIRECTORIES.md

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