librelad 6bb04533fa fix(desudo): manager->self sudo drops -> runAsManager (scoped-sudoers safe)
The scoped sudoers grants the manager (root) and (dockerinstall) but NOT
(itself), so the many 'sudo -u $sudo_user_name <cmd>' calls (crontab,
git/update, reinstall, swapfile, …) failed with 'a password is required'
once per CLI command. runAsManager runs the command plainly when already
the manager (the runtime case) and only sudo -u's when root (install
time), so it's correct in both contexts and needs no sudoers self-grant.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Signed-off-by: librelad <librelad@digitalangels.vip>
2026-05-24 18:40:19 +01:00

225 lines
10 KiB
Bash
Executable File

#!/bin/bash
installDockerRootless()
{
if [[ $CFG_DOCKER_INSTALL_TYPE == "rootless" ]]; then
isHeader "Install Docker Rootless"
#dockerComposeDownAllApps root;
#dockerServiceStop root;
((menu_number++))
echo ""
echo "---- $menu_number. Installing System Requirements."
echo ""
local docker_install_user_id=$(id -u "$CFG_DOCKER_INSTALL_USER")
local docker_install_bashrc="/home/$CFG_DOCKER_INSTALL_USER/.bashrc"
local result=$(runSystem apt-get install -y apt-transport-https ca-certificates curl gnupg software-properties-common uidmap dbus-user-session fuse-overlayfs passt)
checkSuccess "Installing necessary packages"
local result=$(runSystem systemctl disable --now docker.service docker.socket)
checkSuccess "Disabling Docker service & Socket"
((menu_number++))
echo ""
echo "---- $menu_number. Installing slirp4netns."
echo ""
# slirp4netns update and install
if ! command -v slirp4netns &> /dev/null; then
isNotice "slirp4netns is not installed. Installing..."
local result=$(runSystem apt-get install -y slirp4netns)
checkSuccess "Installing slirp4netns"
else
isNotice "slirp4netns is already installed"
installed_version=$(slirp4netns --version | awk '{print $2}')
latest_version=$(curl -s https://api.github.com/repos/rootless-containers/slirp4netns/releases/latest | grep tag_name | cut -d '"' -f 4)
if [[ "$installed_version" != "$latest_version" ]]; then
isNotice "slirp4netns version $installed_version is outdated."
isNotice "Installing version $latest_version..."
local result=$(runSystem apt-get update)
checkSuccess "Updating apt packages"
local result=$(runSystem apt-get install -y slirp4netns)
checkSuccess "Installing slirp4netns"
else
isSuccessful "slirp4netns version $installed_version is up to date"
fi
fi
if [[ $(lsb_release -rs) == "10" ]]; then
((menu_number++))
echo ""
echo "---- $menu_number. Updating the sysctl file for Updating Debian 10."
echo ""
if sudo grep -q "kernel.unprivileged_userns_clone=1" $sysctl; then
isNotice "kernel.unprivileged_userns_clone=1 already exists in $sysctl"
else
local result=$(echo "kernel.unprivileged_userns_clone=1" | sudo tee -a $sysctl > /dev/null)
checkSuccess "Adding kernel.unprivileged_userns_clone=1 to $sysctl..."
local result=$(runSystem sysctl --system)
checkSuccess "Running runAsManager sysctl --system..."
fi
fi
((menu_number++))
echo ""
echo "---- $menu_number. Update the .bashrc file."
echo ""
if ! grep -qF "# DOCKER ROOTLESS BASHRC START" "$docker_install_bashrc"; then
local result=$(echo '# DOCKER ROOTLESS BASHRC START' | sudo tee -a "$docker_install_bashrc" > /dev/null)
checkSuccess "Adding rootless header to .bashrc"
local result=$(echo 'export XDG_RUNTIME_DIR=/run/user/${UID}' | sudo tee -a "$docker_install_bashrc" > /dev/null)
checkSuccess "Adding export path to .bashrc"
local result=$(echo 'export PATH=/usr/bin:$PATH' | sudo tee -a "$docker_install_bashrc" > /dev/null)
checkSuccess "Adding export path to .bashrc"
local result=$(echo 'export DOCKER_HOST=unix:///run/user/${UID}/docker.sock' | sudo tee -a "$docker_install_bashrc" > /dev/null)
checkSuccess "Adding export DOCKER_HOST path to .bashrc"
local result=$(echo 'export DBUS_SESSION_BUS_ADDRESS="unix:path=/run/user/${UID}/bus"' | sudo tee -a "$docker_install_bashrc" > /dev/null)
checkSuccess "Adding export DBUS_SESSION_BUS_ADDRESS path to .bashrc"
local result=$(echo '# DOCKER ROOTLESS BASHRC END' | sudo tee -a "$docker_install_bashrc" > /dev/null)
checkSuccess "Adding rootless header to .bashrc"
isSuccessful "Added $CFG_DOCKER_INSTALL_USER to bashrc file"
else
isNotice "Rootless .bashrc already configured for $CFG_DOCKER_INSTALL_USER — skipping"
fi
((menu_number++))
echo ""
echo "---- $menu_number. Setting up Rootless Docker."
echo ""
local result=$(runSystem loginctl enable-linger $CFG_DOCKER_INSTALL_USER)
checkSuccess "Adding automatic start (linger)"
# Rootless Install
rootless_install=$(cat <<EOF
curl -fsSL https://get.docker.com/rootless | sh && \
systemctl --user start docker && \
systemctl --user enable docker && \
exit
EOF
)
local result=$(dockerCommandRunInstallUser "$rootless_install")
checkSuccess "Setting up Rootless for $CFG_DOCKER_INSTALL_USER"
((menu_number++))
# The net namespace driver and the rootlesskit port driver are a matched
# pair — mixing them makes rootlesskit reject the config and the daemon
# silently never starts (this is why an earlier pasta+builtin attempt
# bricked the install). The user picks one network stack via
# CFG_ROOTLESS_NET; we derive its only valid port driver here:
# pasta -> implicit (fastest; propagates the real client IP)
# slirp4netns -> builtin (legacy fallback)
local rootless_net="${CFG_ROOTLESS_NET:-pasta}"
local rootless_port_driver
case "$rootless_net" in
pasta) rootless_port_driver="implicit" ;;
slirp4netns) rootless_port_driver="builtin" ;;
*)
isNotice "Unknown CFG_ROOTLESS_NET='$rootless_net'; falling back to pasta."
rootless_net="pasta"
rootless_port_driver="implicit"
;;
esac
echo ""
echo "---- $menu_number. Configuring rootless networking ($rootless_net + $rootless_port_driver port driver)."
echo ""
systemd_user_dir="/home/$CFG_DOCKER_INSTALL_USER/.config/systemd/user"
local result=$(dockerCommandRunInstallUser "mkdir -p $systemd_user_dir")
checkSuccess "Create the systemd user directory if it doesn't exist"
local result=$(dockerCommandRunInstallUser "mkdir -p $systemd_user_dir/docker.service.d")
checkSuccess "Create the docker.service.d directory if it doesn't exist"
override_conf_file="$systemd_user_dir/docker.service.d/override.conf"
local result=$(sudo touch $override_conf_file)
checkSuccess "Create the override.conf in docker.service.d"
sudo bash -c "cat <<EOL > '$override_conf_file'
[Service]
Environment='DOCKERD_ROOTLESS_ROOTLESSKIT_NET=$rootless_net'
Environment='DOCKERD_ROOTLESS_ROOTLESSKIT_PORT_DRIVER=$rootless_port_driver'
Environment='DOCKERD_ROOTLESS_ROOTLESSKIT_MTU=$CFG_NETWORK_MTU'
EOL"
local result=$(sudo chown $CFG_DOCKER_INSTALL_USER:$CFG_DOCKER_INSTALL_USER $override_conf_file)
checkSuccess "Updating ownership for override.conf"
# NOTE: we deliberately do NOT set "userland-proxy": false here. Disabling
# it makes rootless dockerd require br_netfilter
# (/proc/sys/net/bridge/bridge-nf-call-iptables), which isn't present in
# the rootless netns on Debian — the daemon then fails to create the
# default bridge and won't start. The userland proxy's lack of source-IP
# propagation doesn't matter here: apps sit behind Traefik, which carries
# the real client IP via X-Forwarded-For at L7.
local result=$(dockerCommandRunInstallUser "systemctl --user daemon-reload")
checkSuccess "Reload the systemd user manager configuration"
isNotice "Restarting docker service...this may take a moment..."
local result=$(dockerCommandRunInstallUser "systemctl --user restart docker")
checkSuccess "Reload the systemd user docker service"
local result=$(sudo cp $sysctl $sysctl.bak)
checkSuccess "Backing up sysctl file"
((menu_number++))
echo ""
echo "---- $menu_number. Setting up sysctl file to work with LetsEncrypt."
echo ""
# Update sysctl file
if ! grep -qF "# DOCKER ROOTLESS SYSCTL START" "$sysctl"; then
local result=$(echo '# DOCKER ROOTLESS SYSCTL START' | sudo tee -a "$sysctl" > /dev/null)
checkSuccess "Adding rootless header to sysctl"
local result=$(echo 'net.ipv4.ip_unprivileged_port_start=0' | sudo tee -a "$sysctl" > /dev/null)
checkSuccess "Adding ip_unprivileged_port_start to sysctl"
local result=$(echo 'kernel.unprivileged_userns_clone=1' | sudo tee -a "$sysctl" > /dev/null)
checkSuccess "Adding unprivileged_userns_clone to sysctl"
local result=$(echo '# DOCKER ROOTLESS SYSCTL END' | sudo tee -a "$sysctl" > /dev/null)
checkSuccess "Adding rootless end to sysctl"
isSuccessful "Updated the sysctl with Docker Rootless configuration"
fi
# Enabling unprivileged user namespaces (needed for rootless) widens the
# kernel attack surface reachable by unprivileged users. Offset that by
# closing the surfaces that local-privilege-escalation chains lean on:
# kptr_restrict hides kernel pointers (info-leak primitives), ptrace_scope
# blocks cross-process ptrace (credential theft post-compromise), and
# bpf_jit_harden hardens the JIT against spraying. Written under
# /etc/sysctl.d/ so `sysctl --system` actually loads it — LibrePortal's
# own $sysctl path (/etc/sysctl/…) is non-standard and is NOT read by
# `sysctl --system`.
local hardening_conf="/etc/sysctl.d/99-libreportal-hardening.conf"
sudo bash -c "cat > '$hardening_conf'" <<'EOL'
# LibrePortal kernel LPE-surface hardening (paired with rootless unprivileged userns)
kernel.kptr_restrict=2
kernel.yama.ptrace_scope=1
net.core.bpf_jit_harden=2
EOL
checkSuccess "Writing kernel LPE-surface hardening to $hardening_conf"
local result=$(runSystem sysctl --system)
checkSuccess "Applying changes to sysctl"
menu_number=0
fi
}