FIREJAIL-PROFILE(5)

📦

version 0.9.82.2 · 11 versions

🌸 NAME 🌸

okay so welcome to FIREJAIL-PROFILE(5), bestie — this is the manual page for profile, aka the security profile file syntax, plus all the info you need for building new application profiles. 💅✨

💖 SYNOPSIS 💖

🎀 Using a specific profile 🎀

run it with a filename like this:

firejail --profile=filename.profile

Example:

$ firejail --appimage
--profile=/etc/firejail/kdenlive.profile kdenlive.appimage

or just use a profile name, no cap:

firejail --profile=profile_name

Example:

$ firejail --appimage --profile=kdenlive kdenlive.appimage

🛠️ Building a profile manually 🛠️

start with the template in /usr/share/doc/firejail/profile.template and modify it in a text editor. to integrate the program in your desktop environment, copy the profile file in ~/.config/firejail directory and run "sudo firecfg". periodt. 🫶

🔁 Aliases and redirections 🔁

in some cases the same profile can be used for several applications. one such example is LibreOffice. build a regular profile for the main application, and for the rest use /usr/share/doc/firejail/redirect_alias-profile.template. iconic behavior honestly. 👑

🧚 Running the profile builder 🧚

firejail --build=appname.profile appname

Example:

$ firejail --build=blobby.profile /usr/bin/blobby

run the program in "firejail --build" and try to exercise as many program features as possible. the profile is extracted and saved in the current directory. open it in a text editor and add or remove sandboxing options as necessary. test again after modifying the profile. to integrate the program in your desktop environment copy the profile file in ~/.config/firejail directory and run "sudo firecfg". 💖

✨ DESCRIPTION ✨

several command line options can be passed to the program using profile files. Firejail chooses the profile file as follows:

if a profile file is provided by the user with --profile option, the profile file is loaded. if a profile name is given, it is searched for first in the ~/.config/firejail directory and if not found then in /etc/firejail directory. profile names do not include the .profile suffix. Example:

$ firejail --profile=/home/netblue/icecat.profile /usr/bin/icecat
Reading profile /home/netblue/icecat.profile
[...]

$ firejail --profile=icecat /usr/bin/icecat-wrapper.sh
Reading profile /etc/firejail/icecat.profile
[...]

if a profile file with the same name as the application is present in ~/.config/firejail directory or in /etc/firejail, the profile is loaded. ~/.config/firejail takes precedence over /etc/firejail. Example:

$ firejail /usr/bin/icecat
Command name #icecat#
Found icecat profile in /home/netblue/.config/firejail directory
Reading profile /home/netblue/.config/firejail/icecat.profile
[...]

use a default.profile file if the sandbox is started by a regular user, or a server.profile file if the sandbox is started by root. Firejail looks for these files in ~/.config/firejail directory, followed by /etc/firejail directory. to disable default profile loading, use --noprofile command option. Example:

$ firejail
Reading profile /etc/firejail/default.profile
Parent pid 8553, child pid 8554
Child process initialized
[...]

$ firejail --noprofile
Parent pid 8553, child pid 8554
Child process initialized
[...]

🌱 Templates 🌱

in /usr/share/doc/firejail there are two templates to write new profiles.

profile.template — for regular profiles
redirect_alias-profile.template — for aliasing/redirecting profiles

🧚 Scripting 🧚

Scripting commands:

📁 File and directory names 📁

file and directory names containing spaces are supported. the space character ' ' should not be escaped. 🫶

Example:

blacklist ~/My Virtual Machines

💬 `# this is a comment` 💬

Example:

# disable networking
net none # this command creates an empty network namespace

❓ `?CONDITIONAL: profile line` ❓

conditionally add profile line.

Example:

?HAS_APPIMAGE: whitelist ${HOME}/special/appimage/dir

this example will load the whitelist profile line only if the --appimage option has been specified on the command line. 💅

currently the only conditionals supported this way are HAS_APPIMAGE, HAS_NET, HAS_NODBUS, HAS_NOSOUND, HAS_PRIVATE and HAS_X11. the conditionals ALLOW_TRAY, BROWSER_DISABLE_U2F and BROWSER_ALLOW_DRM can be enabled or disabled globally in Firejail's configuration file.

the profile line may be any profile line that you would normally use in a profile except for "quiet" and "include" lines.

Note: when using one or more conditionals and --profile, it is recommended that the relevant option(s) (such as --appimage) be specified before --profile, so that their respective conditional(s) (such as ?HAS_APPIMAGE) inside of the profile evaluate to true. 🧠✨

📦 `include other.profile` 📦

include other.profile file.

Example:

include /etc/firejail/disable-common.inc

the file name can be prefixed with a macro such as ${HOME} or ${CFG}. ${HOME} is expanded as user home directory, and ${CFG} is expanded as Firejail system configuration directory — in most cases /etc/firejail or /usr/local/etc/firejail.

Example: "include ${HOME}/myprofiles/profile1" will load "~/myprofiles/profile1" file.

Example: "include ${CFG}/firefox.profile" will load "/etc/firejail/firefox.profile" file.

the file name may also be just the name without the leading directory components. in this case, first the user config directory (${HOME}/.config/firejail) is searched for the file name and if not found then the system configuration directory is searched for the file name. Note: unlike the --profile option which takes a profile name without the '.profile' suffix, include must be given the full file name.

Example: "include firefox.profile" will load "${HOME}/.config/firejail/firefox.profile" file and if it does not exist "${CFG}/firefox.profile" will be loaded.

system configuration files in ${CFG} are overwritten during software installation. persistent configuration at system level is handled in ".local" files. for every profile file in ${CFG} directory, the user can create a corresponding .local file storing modifications to the persistent configuration. persistent .local files are included at the start of regular profile files. 🔥

🚫 `noblacklist file_name` 🚫

if the file name matches file_name, the file will not be blacklisted in any blacklist commands that follow.

Example:

noblacklist ${HOME}/.mozilla

🚫 `nowhitelist file_name` 🚫

if the file name matches file_name, the file will not be whitelisted in any whitelist commands that follow.

Example:

nowhitelist ~/.config

🤷 `ignore` 🤷

ignore command.

Example:

ignore seccomp

Example:

ignore net eth0

🤫 `quiet` 🤫

disable Firejail's output. this should be the first uncommented command in the profile file.

Example:

quiet

⚠️ `warn message` ⚠️

print a warning message to stderr. everything after the command is printed in a warning message as is (that is, without macro expansion).

Example:

warn This file is deprecated; use foo instead.

💾 Filesystem 💾

Okay bestie, these profile entries are literally building a chroot filesystem on top of the existing host filesystem. Each line describes a file/directory that is inaccessible (blacklist), a read-only file or directory (read-only), a tmpfs mounted on top of an existing directory (tmpfs), or mount-bind a directory or file on top of another directory or file (bind). Use private to set private mode. File globbing is supported, and PATH and HOME directories are searched — see the firejail(1) FILE GLOBBING section for more details. Examples:

blacklist file_or_directory — Blacklist directory or file. This makes a file or directory completely inaccessible. All other files and directories are unaffected. The blacklisted file or directory is still visible on the filesystem, even if it's inaccessible.

Symbolic link handling: blacklisting a path that is a symbolic link will also blacklist the path that it points to. For example, if ~/foo is blacklisted and it points to /bar, then /bar will also be blacklisted. Symlinks to /dev/null are not followed (only the symlink itself will be affected), as replacing certain paths with symlinks to /dev/null is not uncommon and restricting access to /dev/null can lead to unexpected issues (see #5803). Using /dev/null directly (instead of through a symlink) should work normally.

Examples:

blacklist /usr/bin
blacklist /usr/bin/gcc*
blacklist ${PATH}/ifconfig
blacklist ${HOME}/.ssh

bind directory1,directory2 — Mount-bind directory1 on top of directory2. This option is only available when running as root. 🫶
bind file1,file2 — Mount-bind file1 on top of file2. This option is only available when running as root. 🫶
disable-mnt — Disable /mnt, /media, /run/mount and /run/media access.
keep-config-pulse — Disable automatic ~/.config/pulse init, for complex setups such as remote pulse servers or non-standard socket paths.
keep-dev-ntsync — Allow access to the /dev/ntsync character device (even with private-dev), which is blocked by default. This device is mostly intended to increase performance and compatibility when running certain programs through Wine.
keep-dev-shm — /dev/shm directory is untouched (even with private-dev).
keep-dev-tpm — Allow access to Trusted Cryptography Module (TCM) and Trusted Platform Module (TPM) devices (even with private-dev), which are blocked by default.
keep-shell-rc — Do not copy shell rc files (such as ~/.bashrc and ~/.zshrc) from /etc/skel.
keep-var-tmp — /var/tmp directory is untouched.
mkdir directory — Create a directory in user home, under /tmp, or under /run/user/<UID> before the sandbox is started. The directory is created if it doesn't already exist.

Use this command for whitelisted directories you need to preserve when the sandbox is closed. Without it, the application will create the directory, and the directory will be deleted when the sandbox is closed. Subdirectories are recursively created. Example from firefox profile:

mkdir ~/.mozilla
whitelist ~/.mozilla
mkdir ~/.cache/mozilla/firefox
whitelist ~/.cache/mozilla/firefox

For files in /run/user/<PID> use ${RUNUSER} macro:

mkdir ${RUNUSER}/firejail-testing

mkfile file — Similar to mkdir, this command creates an empty file in user home, or /tmp, or under /run/user/<UID> before the sandbox is started. The file is created if it doesn't already exist.
noexec file_or_directory — Remount the file or the directory noexec, nodev and nosuid.
private — Mount new /root and /home/user directories in temporary filesystems. All modifications are discarded when the sandbox is closed. ✨
private directory — Use directory as user home. --private and --private=directory cannot be used together.

Bug: Even with this enabled, some firejail commands (such as mkdir, mkfile and private-cache) will still operate on the original home directory. Workaround: Disable the incompatible commands, such as by using "ignore mkdir" and "ignore mkfile". For details, see #903.

private-bin file,file — Build a new /bin in a temporary filesystem, and copy the programs in the list. The files in the list must be expressed as relative to the /bin, /sbin, /usr/bin, /usr/sbin, or /usr/local/bin directories. The same directory is also bind-mounted over /sbin, /usr/bin and /usr/sbin. Multiple private-bin commands are allowed and they accumulate.
private-cache — Mount an empty temporary filesystem on top of the .cache directory in user home. All modifications are discarded when the sandbox is closed.
private-cwd — Set working directory inside jail to the home directory, and failing that, the root directory.
private-cwd directory — Set working directory inside the jail. Full directory path is required. Symbolic links are not allowed.
private-dev — Create a new /dev directory. Only disc, dri, dvb, full, hidraw, log, null, ptmx, pts, random, shm, snd, tty, urandom, usb, video and zero devices are available. Use the options no3d, nodvd, nosound, notv, nou2f and novideo for additional restrictions.
private-etc file,directory — Build a new /etc in a temporary filesystem, and copy the files and directories in the list. The files and directories in the list must be expressed as relative to the /etc directory, and must not contain the / character (e.g., /etc/foo must be expressed as foo, but /etc/foo/bar — expressed as foo/bar — is disallowed). All modifications are discarded when the sandbox is closed. Multiple private-etc commands are allowed and they accumulate.
private-opt file,directory — Build a new /opt in a temporary filesystem, and copy the files and directories in the list. The files and directories in the list must be expressed as relative to the /opt directory, and must not contain the / character (e.g., /opt/foo must be expressed as foo, but /opt/foo/bar — expressed as foo/bar — is disallowed). All modifications are discarded when the sandbox is closed.
private-srv file,directory — Build a new /srv in a temporary filesystem, and copy the files and directories in the list. The files and directories in the list must be expressed as relative to the /srv directory, and must not contain the / character (e.g., /srv/foo must be expressed as foo, but /srv/foo/bar — expressed as foo/bar — is disallowed). All modifications are discarded when the sandbox is closed.
private-tmp — Mount an empty temporary filesystem on top of /tmp directory whitelisting /tmp/.X11-unix.
read-only file_or_directory — Make directory or file read-only.
read-write file_or_directory — Make directory or file read-write.
tmpfs directory — Mount an empty tmpfs filesystem on top of directory. Directories outside user home or not owned by the user are not allowed. Sandboxes running as root are exempt from these restrictions.
whitelist file_or_directory — Whitelist directory or file. A temporary file system is mounted on the top directory. In the context of firejail, top directory means, if the whitelisted file's path is for example /etc/somedir/somefile, then the top directory would be /etc. All other top directories like /opt, /usr and so on, haven't changed, so all files there are still accessible, unless a file or directory inside them is also whitelisted. This is why sometimes it's beneficial to use blacklist in combination with whitelist, if used for different top directories.

The whitelisted files are mount-binded inside. Modifications to whitelisted files are persistent, everything else in the same top directory is discarded when the sandbox is closed.

The top directory can be most directories in /, but there are some special cases. The /proc and /sys top directories aren't allowed, but /sys/module is allowed. Also /run/user/$UID, $HOME and all directories in /usr are treated as a top directory.

Symbolic link handling: whitelisting a path that is a symbolic link will also whitelist the path that it points to. For example, if ~/foo is whitelisted and it points to ~/bar, then ~/bar will also be whitelisted.

Restrictions: With the exception of the user home directory, both the link and the real file should be in the same top directory. For symbolic links in the user home directory, both the link and the real file should be owned by the user.

whitelist-ro file_or_directory — Equivalent to whitelist file_or_directory followed by read-only file_or_directory.
writable-etc — Mount /etc directory read-write.
writable-run-user — Disable the default blacklisting of run/user/$UID/systemd and /run/user/$UID/gnupg.
writable-var — Mount /var directory read-write.
writable-var-log — Use the real /var/log directory, not a clone. By default, a tmpfs is mounted on top of /var/log directory, and a skeleton filesystem is created based on the original /var/log.

🔒 Security filters 🔒

okay bestie, this is the part where firejail stops being a cute little folder-shuffler and starts being an actual bouncer for your app 💅 These are the security filters that exist right now, no cap:

allow-bwrap — lets you run a bubblewrap sandbox inside the Firejail sandbox 🪆 fr fr, by default Firejail straight up replaces the /usr/bin/bwrap executable with a do-nothing program, and this says "no, actually, let it live."
allow-debuggers — lets tools like strace and gdb work inside the sandbox by whitelisting the ptrace and process_vm_readv system calls. Debugging rights, granted 👑
apparmor — turns on AppArmor confinement using the "firejail-default" AppArmor profile.
apparmor profile_name — turns on AppArmor confinement with a custom profile of your choosing. Heads up though: that profile has to already be loaded into the kernel, no take-backs 🫶
caps — enables the default Linux capabilities filter. Peep capabilities(7) for the deets.
caps.drop capability,capability,capability — blacklists the capabilities you list. Bye 👋
caps.drop all — blacklists all the Linux capabilities. absolutely zero chill, iconic 💀
caps.keep capability,capability,capability — whitelists the capabilities you list (the opposite vibe of drop).
landlock.enforce (experimental) — actually enforces the Landlock ruleset. And like… without this one, all your other Landlock commands literally do nothing?? so don't skip it 😭
landlock.fs.read path (experimental) — creates a Landlock ruleset (if one doesn't already exist) and adds a read access rule for path.
landlock.fs.write path (experimental) — creates a Landlock ruleset (if one doesn't already exist) and adds a write access rule for path.
landlock.fs.makeipc path (experimental) — creates a Landlock ruleset (if one doesn't already exist) and adds a rule that allows creating named pipes (FIFOs) and Unix domain sockets underneath the given path.
landlock.fs.makedev path (experimental) — creates a Landlock ruleset (if one doesn't already exist) and adds a rule that allows creating block devices and character devices underneath the given path.
landlock.fs.execute path (experimental) — creates a Landlock ruleset (if one doesn't already exist) and adds an execution permission rule for path.
memory-deny-write-execute — installs a seccomp filter that blocks attempts to create memory mappings that are both writable and executable, to change mappings so they become executable, or to create executable shared memory. no RWX shenanigans allowed 🚫
nonewprivs — sets the NO_NEW_PRIVS prctl. This makes sure child processes can't acquire new privileges via execve(2); specifically, it means that calling a suid binary (or one with file capabilities) does not result in an increase of privilege. slay, stay in your lane ✨
noprinters — disables printers. 🖨️❌
noroot — use this one to enable a user namespace. The namespace has only one user — the current user. There's no root account (uid 0) defined in the namespace at all. main character only 💅
protocol protocol1,protocol2,protocol3 — enables the protocol filter. It's based on seccomp and checks the first argument to the socket system call. Recognized values are: unix, inet, inet6, netlink, packet, and bluetooth. You can use multiple protocol commands and they accumulate. stack 'em 🧚
restrict-namespaces — installs a seccomp filter that blocks attempts to create new cgroup, ipc, net, mount, pid, time, user or uts namespaces.
restrict-namespaces cgroup,ipc,net,mnt,pid,time,user,uts — installs a seccomp filter that blocks attempts to create any of the namespaces you specify. (yes, mnt is how you spell it in the list 👀)
seccomp — enables the seccomp filter and blacklists the syscalls in the default list. See man 1 firejail for more.
seccomp.32 — enables the seccomp filter and blacklists the syscalls in the default list for 32-bit system calls on a 64-bit architecture system.
seccomp syscall,syscall,syscall — enables the seccomp filter and blacklists the system calls in your list on top of the default seccomp filter.
seccomp.32 syscall,syscall,syscall — enables the seccomp filter and blacklists the system calls in your list on top of the default seccomp filter, for 32-bit system calls on a 64-bit architecture system.
seccomp.block-secondary — enables the seccomp filter and filters system call architectures so that only the native architecture is allowed. one architecture, one love 🫶
seccomp.drop syscall,syscall,syscall — enables the seccomp filter and blacklists the system calls in the list.
seccomp.32.drop syscall,syscall,syscall — enables the seccomp filter and blacklists the system calls in the list for 32-bit system calls on a 64-bit architecture system.
seccomp.keep syscall,syscall,syscall — enables the seccomp filter and whitelists the system calls in the list.
seccomp.32.keep syscall,syscall,syscall — enables the seccomp filter and whitelists the system calls in the list for 32-bit system calls on a 64-bit architecture system.
seccomp-error-action kill | log | ERRNO — return a different error instead of EPERM to the process, kill it when an attempt is made to call a blocked system call, or allow but log the attempt. choose your fighter ⚔️
x11 — enables X11 sandboxing.
x11 none — blacklists the /tmp/.X11-unix directory, ${HOME}/.Xauthority and the file specified in the ${XAUTHORITY} environment variable. It also removes the DISPLAY and XAUTHORITY environment variables. And it stops with an error message if the X11 abstract socket would be accessible inside the jail 😤 no sneaky sockets.
xephyr-extra-params OPTIONS — sets extra parameters for the Xephyr server command used by x11. This command should be included in the profile file before the x11 command.

Example:

xephyr-extra-params -grayscale
x11

xephyr-screen WIDTHxHEIGHT — sets the screen size for the x11 command. This one should also be included in the profile file before the x11 command.

Example:

xephyr-screen 640x480
x11

🌈 DBus filtering 🌈

okay so access to the session and system DBus UNIX sockets can be allowed, filtered, or disabled — three whole moods 💅 To disable the abstract sockets (and force applications to use the filtered UNIX socket) you'd need to request a new network namespace using the --net command. Another option is to remove unix from the --protocol set. iykyk 🦋

Filtering requires installing the xdg-dbus-proxy utility. Filter rules can be specified for well-known DBus names, but they also get propagated to the owning unique name, too. The permissions are "sticky" and are kept even if the corresponding well-known name is released (though like… applications rarely release well-known names in practice 🙈). Names may have a .* suffix to match all names underneath them, including themselves — e.g. "foo.bar.*" matches "foo.bar", "foo.bar.baz" and "foo.bar.baz.quux", but not "foobar". For more info, see xdg-dbus-proxy(1).

Examples, bestie 🎀:

dbus-system filter — enables filtered access to the system DBus. Filters can be specified with the dbus-system.talk and dbus-system.own commands.
dbus-system none — disables access to the system DBus. And once access is disabled, it cannot be relaxed back to filtering. one-way door, no cap 🚪
dbus-system.broadcast org.freedesktop.Notifications=org.freedesktop.Notifications.*@/org/freedesktop/Notifications — allows the application to receive broadcast signals from the interface org.freedesktop.Notifications of the object exposed at the path /org/freedesktop/Notifications by the client owning the bus name org.freedesktop.Notifications on the system DBus.
dbus-system.call org.freedesktop.Notifications=org.freedesktop.Notifications.*@/org/freedesktop/Notifications — allows the application to call methods of the interface org.freedesktop.Notifications of the object exposed at the path /org/freedesktop/Notifications by the client owning the bus name org.freedesktop.Notifications on the system DBus.
dbus-system.own org.gnome.ghex.* — allows the application to own the name org.gnome.ghex and all names underneath it on the system DBus. 👑
dbus-system.see org.freedesktop.Notifications — allows the application to see but not talk to the name org.freedesktop.Notifications on the system DBus. look, don't touch 👀
dbus-system.talk org.freedesktop.Notifications — allows the application to talk to the name org.freedesktop.Notifications on the system DBus.
dbus-user filter — enables filtered access to the session DBus. Filters can be specified with the dbus-user.talk and dbus-user.own commands.
dbus-user none — disables access to the session DBus. Once access is disabled, it cannot be relaxed to filtering. same one-way door as above 💀
dbus-user.broadcast org.freedesktop.Notifications=org.freedesktop.Notifications.*@/org/freedesktop/Notifications — allows the application to receive broadcast signals from the interface org.freedesktop.Notifications of the object exposed at the path /org/freedesktop/Notifications by the client owning the bus name org.freedesktop.Notifications on the session DBus.
dbus-user.call org.freedesktop.Notifications=org.freedesktop.Notifications.*@/org/freedesktop/Notifications — allows the application to call methods of the interface org.freedesktop.Notifications of the object exposed at the path /org/freedesktop/Notifications by the client owning the bus name org.freedesktop.Notifications on the session DBus.
dbus-user.own org.gnome.ghex.* — allows the application to own the name org.gnome.ghex and all names underneath it on the session DBus.
dbus-user.see org.freedesktop.Notifications — allows the application to see but not talk to the name org.freedesktop.Notifications on the session DBus.
dbus-user.talk org.freedesktop.Notifications — allows the application to talk to the name org.freedesktop.Notifications on the session DBus.
nodbus (deprecated) — disables D-Bus access (both the system and session buses). Equivalent to dbus-system none and dbus-user none. it's giving legacy, but it's still here 🫶

Individual filters can be overridden via the --ignore command. So like, supposing a profile has:

[...]
dbus-user filter
dbus-user.own org.mozilla.firefox.*
dbus-user.talk org.freedesktop.Notifications
dbus-system none
[...]

and the user wants to disable notifications 😤 — that can be achieved by putting the below in a local override file:

[...]
ignore dbus-user.talk org.freedesktop.Notifications
[...]

and boom, notifications handled 💅

🌸 Resource limits, CPU affinity 🌸

These profile entries are literally the vibe check for how much of your system the sandboxed processes are allowed to eat 💅 They set the limits on system resources (rlimits) for the processes inside the sandbox. Fun fact: the limits can be modified inside the sandbox using the regular ulimit command, so it's not a total lockdown. And the cpu command configures the CPU cores available 👑

Examples, bestie — memorise these:

cpu 0,1,2 — use only CPU cores 0, 1 and 2. That's it, that's the core group chat.
nice -5 — set a nice value of -5 to all processes running inside the sandbox.
rlimit-as 123456789012 — set the maximum size of the process's virtual memory to 123456789012 bytes.
rlimit-cpu 123 — set the maximum CPU time in seconds.
rlimit-fsize 1024 — set the maximum file size that can be created by a process to 1024 bytes.
rlimit-nproc 1000 — set the maximum number of processes that can be created for the real user ID of the calling process to 1000.
rlimit-nofile 500 — set the maximum number of files that can be opened by a process to 500.
rlimit-sigpending 200 — set the maximum number of processes that can be created for the real user ID of the calling process to 200.
timeout hh:mm:ss — kill the sandbox automatically after the time has elapsed 😭 The time is specified in hours/minutes/seconds format. So timeout 00:10:00 is your "go to bed" alarm, fr fr.

🦋 User Environment 🦋

allusers — all user home directories are visible inside the sandbox. By default, only the current user home directory is visible. Like, it's giving "everyone's invited to the slumber party".
env name=value — set environment variable. Examples, slay:

env LD_LIBRARY_PATH=/opt/test/lib
env CFLAGS="-W -Wall -Werror"

ipc-namespace — enable a new IPC namespace if the sandbox was started as a regular user. IPC namespace is enabled by default for sandboxes started as root.

⭐ Note: this only affects the IPC resources that are mentioned in ipc_namespaces(7). It does not affect other IPC resources, such as Unix sockets (see unix(7)). Important caveat, don't @ me.

keep-fd — inherit open file descriptors to sandbox.
name sandboxname — set sandbox name. Example:

name browser

no3d — disable 3D hardware acceleration.
noautopulse (deprecated) — see keep-config-pulse. It's giving legacy energy 🙈
nodvd — disable DVD and audio CD devices.
nogroups — disable supplementary user groups.
noinput — disable input devices.
nosound — disable sound system.
notpm (deprecated) — ignored for compatibility. TPM devices are now blocked by default, see keep-dev-tpm. So this one's just here being iconic and doing nothing.
notv — disable DVB (Digital Video Broadcasting) TV devices.
nou2f — disable U2F devices.
novideo — disable video capture devices.
machine-id — spoof id number in /etc/machine-id file — a new random id is generated inside the sandbox 🎭 main character energy, no cap.

🌈 Networking 🌈

Networking features available in profile files. This is where it gets spicy 🔥

defaultgw address — use this address as default gateway in the new network namespace.
dns address — set a DNS server for the sandbox. Up to three DNS servers can be defined.
hostname name — set a hostname for the sandbox.
hostname-randomize — set sandbox hostname to a random value generated by firejail.
hosts-file file — use file as /etc/hosts.
ip address — assign IP addresses to the last network interface defined by a net command. A default gateway is assigned by default. Example:

net eth0
ip 10.10.20.56

ip none — no IP address and no default gateway are configured for the last interface defined by a net command. Example:

net eth0
ip none

ip6 address — assign IPv6 addresses to the last network interface defined by a net command. Example:

net eth0
ip6 2001:0db8:0:f101::1/64

iprange address,address — assign an IP address in the provided range to the last network interface defined by a net command. A default gateway is assigned by default. Example:

net eth0
iprange 192.168.1.150,192.168.1.160

mac address — assign MAC addresses to the last network interface defined by a net command.
mtu number — assign a MTU value to the last network interface defined by a net command.
net bridge_interface — enable a new network namespace and connect it to this bridge interface. Unless specified with option --ip and --defaultgw, an IP address and a default gateway will be assigned automatically to the sandbox. The IP address is verified using ARP before assignment. The address configured as default gateway is the bridge device IP address. Up to four --net bridge devices can be defined. Mixing bridge and macvlan devices is allowed 💖
net ethernet_interface|wireless_interface — enable a new network namespace and connect it to this ethernet interface using the standard Linux macvlan or ipvlan driver. Unless specified with option --ip and --defaultgw, an IP address and a default gateway will be assigned automatically to the sandbox. The IP address is verified using ARP before assignment. The address configured as default gateway is the default gateway of the host. Up to four --net devices can be defined. Mixing bridge and macvlan devices is allowed.
net none — enable a new, unconnected network namespace. The only interface available in the new namespace is a new loopback interface (lo). Use this option to deny network access to programs that don't really need network access. Introvert mode 🫶
net tap_interface — enable a new network namespace and connect it to this ethernet tap interface using the standard Linux macvlan driver. If the tap interface is not configured, the sandbox will not try to configure the interface inside the sandbox. Please use ip, netmask and defaultgw to specify the configuration.
netfilter — if a new network namespace is created, enabled default network filter.
netfilter filename — if a new network namespace is created, enabled the network filter in filename.
netlock — generate a custom network filter and enable it.
netmask address — use this option when you want to assign an IP address in a new namespace and the parent interface specified by --net is not configured. An IP address and a default gateway address also have to be added.
netns namespace — run the program in a named, persistent network namespace. These can be created and configured using "ip netns".
veth-name name — use this name for the interface connected to the bridge for --net=bridge_interface commands, instead of the default one.

✨ Other ✨

deterministic-exit-code — always exit firejail with the first child's exit status. The default behavior is to use the exit status of the final child to exit, which can be nondeterministic. If you hate surprises, this one's for you 😌
deterministic-shutdown — always shut down the sandbox after the first child has terminated. The default behavior is to keep the sandbox alive as long as it contains running processes.
join-or-start sandboxname — join the sandbox identified by name or start a new one. Same as "firejail --join=sandboxname" command if sandbox with specified name exists, otherwise same as "name sandboxname".

🎀 FILES 🎀

/etc/firejail/appname.profile — global Firejail configuration consisting mainly of profiles for each application supported by default.
$HOME/.config/firejail/appname.profile — user application profiles, will take precedence over the global profiles 👑 (user wins, always).
/usr/share/doc/firejail/profile.template — template for building new profiles.
/usr/share/doc/firejail/redirect_alias-profile.template — template for aliasing/redirecting profiles.

💜 LICENSE 💜

Firejail is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

Homepage: https://firejail.wordpress.com

https://github.com/netblue30/firejail/wiki/Creating-Profiles

Okay that's the whole page bestie, you're officially a firejail profile sommelier now 🍓 This lil masterpiece is FIREJAIL-PROFILE(5), version 0.9.82.2, dated Nov 2026 — go forth and write profiles that absolutely ate, no cap 💅✨

📦 where this page comes from

arch rolling: firejail 0.9.82.2-1 · GPL-2.0-only
💖 girlified by deepseek-ai/DeepSeek-V4.1-Flash-1017 on 2026-11-04: a language model's rewrite of the original, under its terms

FIREJAIL-PROFILE(5)

🌸 NAME 🌸

💖 SYNOPSIS 💖

🎀 Using a specific profile 🎀

🛠️ Building a profile manually 🛠️

🔁 Aliases and redirections 🔁

🧚 Running the profile builder 🧚

✨ DESCRIPTION ✨

🌱 Templates 🌱

🧚 Scripting 🧚

📁 File and directory names 📁

💬 # this is a comment 💬

❓ ?CONDITIONAL: profile line ❓

📦 include other.profile 📦

🚫 noblacklist file_name 🚫

🚫 nowhitelist file_name 🚫

🤷 ignore 🤷

🤫 quiet 🤫

⚠️ warn message ⚠️

💾 Filesystem 💾

🔒 Security filters 🔒

🌈 DBus filtering 🌈

🌸 Resource limits, CPU affinity 🌸

🦋 User Environment 🦋

🌈 Networking 🌈

✨ Other ✨

🎀 FILES 🎀

💜 LICENSE 💜

⭐ SEE ALSO ⭐

💬 `# this is a comment` 💬

❓ `?CONDITIONAL: profile line` ❓

📦 `include other.profile` 📦

🚫 `noblacklist file_name` 🚫

🚫 `nowhitelist file_name` 🚫

🤷 `ignore` 🤷

🤫 `quiet` 🤫

⚠️ `warn message` ⚠️