nixpkgs/nixos/modules/security/acme.nix

771 lines
28 KiB
Nix
Raw Normal View History

{ config, lib, pkgs, options, ... }:
with lib;
let
2015-12-11 17:42:17 +01:00
cfg = config.security.acme;
# Used to calculate timer accuracy for coalescing
numCerts = length (builtins.attrNames cfg.certs);
_24hSecs = 60 * 60 * 24;
# Used to make unique paths for each cert/account config set
mkHash = with builtins; val: substring 0 20 (hashString "sha256" val);
mkAccountHash = acmeServer: data: mkHash "${toString acmeServer} ${data.keyType} ${data.email}";
accountDirRoot = "/var/lib/acme/.lego/accounts/";
# There are many services required to make cert renewals work.
# They all follow a common structure:
# - They inherit this commonServiceConfig
# - They all run as the acme user
# - They all use BindPath and StateDirectory where possible
# to set up a sort of build environment in /tmp
# The Group can vary depending on what the user has specified in
# security.acme.certs.<cert>.group on some of the services.
commonServiceConfig = {
Type = "oneshot";
User = "acme";
Group = mkDefault "acme";
UMask = 0022;
StateDirectoryMode = 750;
ProtectSystem = "full";
PrivateTmp = true;
WorkingDirectory = "/tmp";
};
# In order to avoid race conditions creating the CA for selfsigned certs,
# we have a separate service which will create the necessary files.
selfsignCAService = {
description = "Generate self-signed certificate authority";
path = with pkgs; [ minica ];
unitConfig = {
ConditionPathExists = "!/var/lib/acme/.minica/key.pem";
};
serviceConfig = commonServiceConfig // {
StateDirectory = "acme/.minica";
BindPaths = "/var/lib/acme/.minica:/tmp/ca";
};
# Working directory will be /tmp
script = ''
minica \
--ca-key ca/key.pem \
--ca-cert ca/cert.pem \
--domains selfsigned.local
chmod 600 ca/*
'';
};
# Ensures that directories which are shared across all certs
# exist and have the correct user and group, since group
# is configurable on a per-cert basis.
userMigrationService = let
script = with builtins; ''
chown -R acme .lego/accounts
'' + (concatStringsSep "\n" (mapAttrsToList (cert: data: ''
for fixpath in ${escapeShellArg cert} .lego/${escapeShellArg cert}; do
if [ -d "$fixpath" ]; then
chmod -R u=rwX,g=rX,o= "$fixpath"
chown -R acme:${data.group} "$fixpath"
fi
done
'') certConfigs));
in {
description = "Fix owner and group of all ACME certificates";
serviceConfig = commonServiceConfig // {
# We don't want this to run every time a renewal happens
RemainAfterExit = true;
# These StateDirectory entries negate the need for tmpfiles
StateDirectory = [ "acme" "acme/.lego" "acme/.lego/accounts" ];
StateDirectoryMode = 755;
WorkingDirectory = "/var/lib/acme";
# Run the start script as root
ExecStart = "+" + (pkgs.writeShellScript "acme-fixperms" script);
};
};
certToConfig = cert: data: let
acmeServer = if data.server != null then data.server else cfg.server;
useDns = data.dnsProvider != null;
destPath = "/var/lib/acme/${cert}";
selfsignedDeps = optionals (cfg.preliminarySelfsigned) [ "acme-selfsigned-${cert}.service" ];
# Minica and lego have a "feature" which replaces * with _. We need
# to make this substitution to reference the output files from both programs.
# End users never see this since we rename the certs.
keyName = builtins.replaceStrings ["*"] ["_"] data.domain;
# FIXME when mkChangedOptionModule supports submodules, change to that.
# This is a workaround
extraDomains = data.extraDomainNames ++ (
optionals
(data.extraDomains != "_mkMergedOptionModule")
(builtins.attrNames data.extraDomains)
);
# Create hashes for cert data directories based on configuration
# Flags are separated to avoid collisions
hashData = with builtins; ''
${concatStringsSep " " data.extraLegoFlags} -
${concatStringsSep " " data.extraLegoRunFlags} -
${concatStringsSep " " data.extraLegoRenewFlags} -
${toString acmeServer} ${toString data.dnsProvider}
${toString data.ocspMustStaple} ${data.keyType}
'';
certDir = mkHash hashData;
domainHash = mkHash "${concatStringsSep " " extraDomains} ${data.domain}";
accountHash = (mkAccountHash acmeServer data);
accountDir = accountDirRoot + accountHash;
protocolOpts = if useDns then (
[ "--dns" data.dnsProvider ]
++ optionals (!data.dnsPropagationCheck) [ "--dns.disable-cp" ]
++ optionals (data.dnsResolver != null) [ "--dns.resolvers" data.dnsResolver ]
) else (
[ "--http" "--http.webroot" data.webroot ]
);
commonOpts = [
"--accept-tos" # Checking the option is covered by the assertions
"--path" "."
"-d" data.domain
"--email" data.email
"--key-type" data.keyType
] ++ protocolOpts
++ optionals (acmeServer != null) [ "--server" acmeServer ]
++ concatMap (name: [ "-d" name ]) extraDomains
++ data.extraLegoFlags;
# Although --must-staple is common to both modes, it is not declared as a
# mode-agnostic argument in lego and thus must come after the mode.
runOpts = escapeShellArgs (
commonOpts
++ [ "run" ]
++ optionals data.ocspMustStaple [ "--must-staple" ]
++ data.extraLegoRunFlags
);
renewOpts = escapeShellArgs (
commonOpts
++ [ "renew" "--reuse-key" ]
++ optionals data.ocspMustStaple [ "--must-staple" ]
++ data.extraLegoRenewFlags
);
in {
inherit accountHash cert selfsignedDeps;
group = data.group;
renewTimer = {
description = "Renew ACME Certificate for ${cert}";
wantedBy = [ "timers.target" ];
timerConfig = {
OnCalendar = cfg.renewInterval;
Unit = "acme-${cert}.service";
Persistent = "yes";
# Allow systemd to pick a convenient time within the day
# to run the check.
# This allows the coalescing of multiple timer jobs.
# We divide by the number of certificates so that if you
# have many certificates, the renewals are distributed over
# the course of the day to avoid rate limits.
AccuracySec = "${toString (_24hSecs / numCerts)}s";
# Skew randomly within the day, per https://letsencrypt.org/docs/integration-guide/.
RandomizedDelaySec = "24h";
};
};
selfsignService = {
description = "Generate self-signed certificate for ${cert}";
after = [ "acme-selfsigned-ca.service" "acme-fixperms.service" ];
2020-09-04 19:48:47 +02:00
requires = [ "acme-selfsigned-ca.service" "acme-fixperms.service" ];
path = with pkgs; [ minica ];
unitConfig = {
ConditionPathExists = "!/var/lib/acme/${cert}/key.pem";
};
serviceConfig = commonServiceConfig // {
Group = data.group;
StateDirectory = "acme/${cert}";
BindPaths = [
"/var/lib/acme/.minica:/tmp/ca"
"/var/lib/acme/${cert}:/tmp/${keyName}"
];
};
# Working directory will be /tmp
# minica will output to a folder sharing the name of the first domain
# in the list, which will be ${data.domain}
script = ''
minica \
--ca-key ca/key.pem \
--ca-cert ca/cert.pem \
--domains ${escapeShellArg (builtins.concatStringsSep "," ([ data.domain ] ++ extraDomains))}
# Create files to match directory layout for real certificates
cd '${keyName}'
cp ../ca/cert.pem chain.pem
cat cert.pem chain.pem > fullchain.pem
cat key.pem fullchain.pem > full.pem
chmod 640 *
# Group might change between runs, re-apply it
chown 'acme:${data.group}' *
'';
};
renewService = {
description = "Renew ACME certificate for ${cert}";
after = [ "network.target" "network-online.target" "acme-fixperms.service" "nss-lookup.target" ] ++ selfsignedDeps;
wants = [ "network-online.target" "acme-fixperms.service" ] ++ selfsignedDeps;
# https://github.com/NixOS/nixpkgs/pull/81371#issuecomment-605526099
wantedBy = optionals (!config.boot.isContainer) [ "multi-user.target" ];
path = with pkgs; [ lego coreutils diffutils openssl ];
serviceConfig = commonServiceConfig // {
Group = data.group;
# Keep in mind that these directories will be deleted if the user runs
# systemctl clean --what=state
# acme/.lego/${cert} is listed for this reason.
StateDirectory = [
"acme/${cert}"
"acme/.lego/${cert}"
"acme/.lego/${cert}/${certDir}"
"acme/.lego/accounts/${accountHash}"
];
# Needs to be space separated, but can't use a multiline string because that'll include newlines
BindPaths = [
"${accountDir}:/tmp/accounts"
"/var/lib/acme/${cert}:/tmp/out"
"/var/lib/acme/.lego/${cert}/${certDir}:/tmp/certificates"
];
# Only try loading the credentialsFile if the dns challenge is enabled
EnvironmentFile = mkIf useDns data.credentialsFile;
2020-09-04 19:48:47 +02:00
# Run as root (Prefixed with +)
ExecStartPost = "+" + (pkgs.writeShellScript "acme-postrun" ''
cd /var/lib/acme/${escapeShellArg cert}
if [ -e renewed ]; then
rm renewed
${data.postRun}
fi
'');
};
# Working directory will be /tmp
script = ''
set -euxo pipefail
# This reimplements the expiration date check, but without querying
# the acme server first. By doing this offline, we avoid errors
# when the network or DNS are unavailable, which can happen during
# nixos-rebuild switch.
is_expiration_skippable() {
pem=$1
# This function relies on set -e to exit early if any of the
# conditions or programs fail.
[[ -e $pem ]]
expiration_line="$(
set -euxo pipefail
openssl x509 -noout -enddate <$pem \
| grep notAfter \
| sed -e 's/^notAfter=//'
)"
[[ -n "$expiration_line" ]]
expiration_date="$(date -d "$expiration_line" +%s)"
now="$(date +%s)"
expiration_s=$[expiration_date - now]
expiration_days=$[expiration_s / (3600 * 24)] # rounds down
[[ $expiration_days -gt ${toString cfg.validMinDays} ]]
}
${optionalString (data.webroot != null) ''
# Ensure the webroot exists. Fixing group is required in case configuration was changed between runs.
# Lego will fail if the webroot does not exist at all.
(
mkdir -p '${data.webroot}/.well-known/acme-challenge' \
&& chgrp '${data.group}' ${data.webroot}/.well-known/acme-challenge
) || (
echo 'Please ensure ${data.webroot}/.well-known/acme-challenge exists and is writable by acme:${data.group}' \
&& exit 1
)
''}
echo '${domainHash}' > domainhash.txt
# Check if we can renew
if [ -e 'certificates/${keyName}.key' -a -e 'certificates/${keyName}.crt' -a -n "$(ls -1 accounts)" ]; then
# When domains are updated, there's no need to do a full
# Lego run, but it's likely renew won't work if days is too low.
if [ -e certificates/domainhash.txt ] && cmp -s domainhash.txt certificates/domainhash.txt; then
if is_expiration_skippable out/full.pem; then
echo 1>&2 "nixos-acme: skipping renewal because expiration isn't within the coming ${toString cfg.validMinDays} days"
else
echo 1>&2 "nixos-acme: renewing now, because certificate expires within the configured ${toString cfg.validMinDays} days"
lego ${renewOpts} --days ${toString cfg.validMinDays}
fi
else
echo 1>&2 "certificate domain(s) have changed; will renew now"
# Any number > 90 works, but this one is over 9000 ;-)
lego ${renewOpts} --days 9001
fi
# Otherwise do a full run
else
lego ${runOpts}
fi
mv domainhash.txt certificates/
chmod 640 certificates/*
chmod -R u=rwX,g=,o= accounts/*
# Group might change between runs, re-apply it
chown 'acme:${data.group}' certificates/*
# Copy all certs to the "real" certs directory
CERT='certificates/${keyName}.crt'
if [ -e "$CERT" ] && ! cmp -s "$CERT" out/fullchain.pem; then
2020-09-04 19:48:47 +02:00
touch out/renewed
echo Installing new certificate
cp -vp 'certificates/${keyName}.crt' out/fullchain.pem
cp -vp 'certificates/${keyName}.key' out/key.pem
cp -vp 'certificates/${keyName}.issuer.crt' out/chain.pem
ln -sf fullchain.pem out/cert.pem
cat out/key.pem out/fullchain.pem > out/full.pem
fi
'';
};
};
certConfigs = mapAttrs certToConfig cfg.certs;
certOpts = { name, ... }: {
options = {
# user option has been removed
user = mkOption {
visible = false;
default = "_mkRemovedOptionModule";
};
# allowKeysForGroup option has been removed
allowKeysForGroup = mkOption {
visible = false;
default = "_mkRemovedOptionModule";
};
# extraDomains was replaced with extraDomainNames
extraDomains = mkOption {
visible = false;
default = "_mkMergedOptionModule";
};
webroot = mkOption {
type = types.nullOr types.str;
default = null;
example = "/var/lib/acme/acme-challenge";
description = ''
Where the webroot of the HTTP vhost is located.
<filename>.well-known/acme-challenge/</filename> directory
will be created below the webroot if it doesn't exist.
<literal>http://example.org/.well-known/acme-challenge/</literal> must also
be available (notice unencrypted HTTP).
'';
};
server = mkOption {
type = types.nullOr types.str;
default = null;
description = ''
2020-04-21 15:39:46 +02:00
ACME Directory Resource URI. Defaults to Let's Encrypt's
production endpoint,
2020-04-21 15:39:46 +02:00
<link xlink:href="https://acme-v02.api.letsencrypt.org/directory"/>, if unset.
'';
};
domain = mkOption {
type = types.str;
default = name;
2020-04-21 15:39:46 +02:00
description = "Domain to fetch certificate for (defaults to the entry name).";
};
email = mkOption {
type = types.nullOr types.str;
default = cfg.email;
description = "Contact email address for the CA to be able to reach you.";
};
group = mkOption {
type = types.str;
default = "acme";
2015-12-11 17:42:17 +01:00
description = "Group running the ACME client.";
};
postRun = mkOption {
type = types.lines;
default = "";
example = "cp full.pem backup.pem";
description = ''
Commands to run after new certificates go live. Note that
2020-09-04 19:48:47 +02:00
these commands run as the root user.
Executed in the same directory with the new certificate.
'';
};
Fix letsencrypt (#60219) * nixos/acme: Fix ordering of cert requests When subsequent certificates would be added, they would not wake up nginx correctly due to target units only being triggered once. We now added more fine-grained systemd dependencies to make sure nginx always is aware of new certificates and doesn't restart too early resulting in a crash. Furthermore, the acme module has been refactored. Mostly to get rid of the deprecated PermissionStartOnly systemd options which were deprecated. Below is a summary of changes made. * Use SERVICE_RESULT to determine status This was added in systemd v232. we don't have to keep track of the EXITCODE ourselves anymore. * Add regression test for requesting mutliple domains * Deprecate 'directory' option We now use systemd's StateDirectory option to manage create and permissions of the acme state directory. * The webroot is created using a systemd.tmpfiles.rules rule instead of the preStart script. * Depend on certs directly By getting rid of the target units, we make sure ordering is correct in the case that you add new certs after already having deployed some. Reason it broke before: acme-certificates.target would be in active state, and if you then add a new cert, it would still be active and hence nginx would restart without even requesting a new cert. Not good! We make the dependencies more fine-grained now. this should fix that * Remove activationDelay option It complicated the code a lot, and is rather arbitrary. What if your activation script takes more than activationDelay seconds? Instead, one should use systemd dependencies to make sure some action happens before setting the certificate live. e.g. If you want to wait until your cert is published in DNS DANE / TLSA, you could create a unit that blocks until it appears in DNS: ``` RequiredBy=acme-${cert}.service After=acme-${cert}.service ExecStart=publish-wait-for-dns-script ```
2019-08-29 16:32:59 +02:00
directory = mkOption {
type = types.str;
readOnly = true;
default = "/var/lib/acme/${name}";
description = "Directory where certificate and other state is stored.";
};
extraDomainNames = mkOption {
type = types.listOf types.str;
default = [];
example = literalExample ''
[
"example.org"
"mydomain.org"
]
'';
description = ''
2020-04-29 21:31:17 +02:00
A list of extra domain names, which are included in the one certificate to be issued.
'';
};
keyType = mkOption {
type = types.str;
nixos/acme: change default keyType to ec256 Previously, the NixOS ACME module defaulted to using P-384 for TLS certificates. I believe that this is a mistake, and that we should use P-256 instead, despite it being theoretically cryptographically weaker. The security margin of a 256-bit elliptic curve cipher is substantial; beyond a certain level, more bits in the key serve more to slow things down than add meaningful protection. It's much more likely that ECDSA will be broken entirely, or some fatal flaw will be found in the NIST curves that makes them all insecure, than that the security margin will be reduced enough to put P-256 at risk but not P-384. It's also inconsistent to target a curve with a 192-bit security margin when our recommended nginx TLS configuration allows 128-bit AES. [This Stack Exchange answer][pornin] by cryptographer Thomas Pornin conveys the general attitude among experts: > Use P-256 to minimize trouble. If you feel that your manhood is > threatened by using a 256-bit curve where a 384-bit curve is > available, then use P-384: it will increases your computational and > network costs (a factor of about 3 for CPU, a few extra dozen bytes > on the network) but this is likely to be negligible in practice (in a > SSL-powered Web server, the heavy cost is in "Web", not "SSL"). [pornin]: https://security.stackexchange.com/a/78624 While the NIST curves have many flaws (see [SafeCurves][safecurves]), P-256 and P-384 are no different in this respect; SafeCurves gives them the same rating. The only NIST curve Bernstein [thinks better of, P-521][bernstein] (see "Other standard primes"), isn't usable for Web PKI (it's [not supported by BoringSSL by default][boringssl] and hence [doesn't work in Chromium/Chrome][chromium], and Let's Encrypt [don't support it either][letsencrypt]). [safecurves]: https://safecurves.cr.yp.to/ [bernstein]: https://blog.cr.yp.to/20140323-ecdsa.html [boringssl]: https://boringssl.googlesource.com/boringssl/+/e9fc3e547e557492316932b62881c3386973ceb2 [chromium]: https://bugs.chromium.org/p/chromium/issues/detail?id=478225 [letsencrypt]: https://letsencrypt.org/docs/integration-guide/#supported-key-algorithms So there's no real benefit to using P-384; what's the cost? In the Stack Exchange answer I linked, Pornin estimates a factor of 3× CPU usage, which wouldn't be so bad; unfortunately, this is wildly optimistic in practice, as P-256 is much more common and therefore much better optimized. [This GitHub comment][openssl] measures the performance differential for raw Diffie-Hellman operations with OpenSSL 1.1.1 at a whopping 14× (even P-521 fares better!); [Caddy disables P-384 by default][caddy] due to Go's [lack of accelerated assembly implementations][crypto/elliptic] for it, and the difference there seems even more extreme: [this golang-nuts post][golang-nuts] measures the key generation performance differential at 275×. It's unlikely to be the bottleneck for anyone, but I still feel kind of bad for anyone having lego generate hundreds of certificates and sign challenges with them with performance like that... [openssl]: https://github.com/mozilla/server-side-tls/issues/190#issuecomment-421831599 [caddy]: https://github.com/caddyserver/caddy/blob/2cab475ba516fa725d012f53ca417c3e039607de/modules/caddytls/values.go#L113-L124 [crypto/elliptic]: https://github.com/golang/go/tree/2910c5b4a01a573ebc97744890a07c1a3122c67a/src/crypto/elliptic [golang-nuts]: https://groups.google.com/forum/#!topic/golang-nuts/nlnJkBMMyzk In conclusion, there's no real reason to use P-384 in general: if you don't care about Web PKI compatibility and want to use a nicer curve, then Ed25519 or P-521 are better options; if you're a NIST-fearing paranoiac, you should use good old RSA; but if you're a normal person running a web server, then you're best served by just using P-256. Right now, NixOS makes an arbitrary decision between two equally-mediocre curves that just so happens to slow down ECDH key agreement for every TLS connection by over an order of magnitude; this commit fixes that. Unfortunately, it seems like existing P-384 certificates won't get migrated automatically on renewal without manual intervention, but that's a more general problem with the existing ACME module (see #81634; I know @yegortimoshenko is working on this). To migrate your certificates manually, run: $ sudo find /var/lib/acme/.lego/certificates -type f -delete $ sudo find /var/lib/acme -name '*.pem' -delete $ sudo systemctl restart 'acme-*.service' nginx.service (No warranty. If it breaks, you get to keep both pieces. But it worked for me.)
2020-03-22 06:27:20 +01:00
default = "ec256";
description = ''
Key type to use for private keys.
For an up to date list of supported values check the --key-type option
2020-04-21 15:39:46 +02:00
at <link xlink:href="https://go-acme.github.io/lego/usage/cli/#usage"/>.
'';
};
dnsProvider = mkOption {
type = types.nullOr types.str;
default = null;
example = "route53";
description = ''
DNS Challenge provider. For a list of supported providers, see the "code"
2020-04-21 15:39:46 +02:00
field of the DNS providers listed at <link xlink:href="https://go-acme.github.io/lego/dns/"/>.
'';
};
dnsResolver = mkOption {
type = types.nullOr types.str;
default = null;
example = "1.1.1.1:53";
description = ''
Set the resolver to use for performing recursive DNS queries. Supported:
host:port. The default is to use the system resolvers, or Google's DNS
resolvers if the system's cannot be determined.
'';
};
credentialsFile = mkOption {
type = types.path;
description = ''
Path to an EnvironmentFile for the cert's service containing any required and
optional environment variables for your selected dnsProvider.
To find out what values you need to set, consult the documentation at
2020-04-21 15:39:46 +02:00
<link xlink:href="https://go-acme.github.io/lego/dns/"/> for the corresponding dnsProvider.
'';
example = "/var/src/secrets/example.org-route53-api-token";
};
dnsPropagationCheck = mkOption {
type = types.bool;
default = true;
description = ''
Toggles lego DNS propagation check, which is used alongside DNS-01
challenge to ensure the DNS entries required are available.
'';
};
2020-02-23 02:51:19 +01:00
ocspMustStaple = mkOption {
type = types.bool;
default = false;
description = ''
Turns on the OCSP Must-Staple TLS extension.
Make sure you know what you're doing! See:
<itemizedlist>
<listitem><para><link xlink:href="https://blog.apnic.net/2019/01/15/is-the-web-ready-for-ocsp-must-staple/" /></para></listitem>
<listitem><para><link xlink:href="https://blog.hboeck.de/archives/886-The-Problem-with-OCSP-Stapling-and-Must-Staple-and-why-Certificate-Revocation-is-still-broken.html" /></para></listitem>
</itemizedlist>
'';
};
2020-06-08 02:17:55 +02:00
extraLegoFlags = mkOption {
type = types.listOf types.str;
default = [];
description = ''
Additional global flags to pass to all lego commands.
'';
};
extraLegoRenewFlags = mkOption {
type = types.listOf types.str;
default = [];
description = ''
Additional flags to pass to lego renew.
'';
};
extraLegoRunFlags = mkOption {
type = types.listOf types.str;
default = [];
description = ''
Additional flags to pass to lego run.
'';
};
};
};
in {
options = {
2015-12-11 17:42:17 +01:00
security.acme = {
validMinDays = mkOption {
type = types.int;
default = 30;
description = "Minimum remaining validity before renewal in days.";
};
email = mkOption {
type = types.nullOr types.str;
default = null;
description = "Contact email address for the CA to be able to reach you.";
};
renewInterval = mkOption {
type = types.str;
default = "daily";
description = ''
Systemd calendar expression when to check for renewal. See
<citerefentry><refentrytitle>systemd.time</refentrytitle>
<manvolnum>7</manvolnum></citerefentry>.
'';
};
server = mkOption {
type = types.nullOr types.str;
default = null;
description = ''
2020-04-21 15:39:46 +02:00
ACME Directory Resource URI. Defaults to Let's Encrypt's
production endpoint,
2020-04-21 15:39:46 +02:00
<link xlink:href="https://acme-v02.api.letsencrypt.org/directory"/>, if unset.
'';
};
preliminarySelfsigned = mkOption {
type = types.bool;
default = true;
description = ''
Whether a preliminary self-signed certificate should be generated before
doing ACME requests. This can be useful when certificates are required in
a webserver, but ACME needs the webserver to make its requests.
With preliminary self-signed certificate the webserver can be started and
can later reload the correct ACME certificates.
'';
};
acceptTerms = mkOption {
type = types.bool;
default = false;
description = ''
2020-04-21 15:39:46 +02:00
Accept the CA's terms of service. The default provider is Let's Encrypt,
you can find their ToS at <link xlink:href="https://letsencrypt.org/repository/"/>.
'';
};
certs = mkOption {
default = { };
type = with types; attrsOf (submodule certOpts);
description = ''
Fix letsencrypt (#60219) * nixos/acme: Fix ordering of cert requests When subsequent certificates would be added, they would not wake up nginx correctly due to target units only being triggered once. We now added more fine-grained systemd dependencies to make sure nginx always is aware of new certificates and doesn't restart too early resulting in a crash. Furthermore, the acme module has been refactored. Mostly to get rid of the deprecated PermissionStartOnly systemd options which were deprecated. Below is a summary of changes made. * Use SERVICE_RESULT to determine status This was added in systemd v232. we don't have to keep track of the EXITCODE ourselves anymore. * Add regression test for requesting mutliple domains * Deprecate 'directory' option We now use systemd's StateDirectory option to manage create and permissions of the acme state directory. * The webroot is created using a systemd.tmpfiles.rules rule instead of the preStart script. * Depend on certs directly By getting rid of the target units, we make sure ordering is correct in the case that you add new certs after already having deployed some. Reason it broke before: acme-certificates.target would be in active state, and if you then add a new cert, it would still be active and hence nginx would restart without even requesting a new cert. Not good! We make the dependencies more fine-grained now. this should fix that * Remove activationDelay option It complicated the code a lot, and is rather arbitrary. What if your activation script takes more than activationDelay seconds? Instead, one should use systemd dependencies to make sure some action happens before setting the certificate live. e.g. If you want to wait until your cert is published in DNS DANE / TLSA, you could create a unit that blocks until it appears in DNS: ``` RequiredBy=acme-${cert}.service After=acme-${cert}.service ExecStart=publish-wait-for-dns-script ```
2019-08-29 16:32:59 +02:00
Attribute set of certificates to get signed and renewed. Creates
<literal>acme-''${cert}.{service,timer}</literal> systemd units for
each certificate defined here. Other services can add dependencies
to those units if they rely on the certificates being present,
or trigger restarts of the service if certificates get renewed.
'';
example = literalExample ''
{
"example.com" = {
webroot = "/var/lib/acme/acme-challenge/";
email = "foo@example.com";
extraDomainNames = [ "www.example.com" "foo.example.com" ];
};
"bar.example.com" = {
webroot = "/var/lib/acme/acme-challenge/";
email = "bar@example.com";
};
}
'';
};
};
};
imports = [
(mkRemovedOptionModule [ "security" "acme" "production" ] ''
Use security.acme.server to define your staging ACME server URL instead.
To use the let's encrypt staging server, use security.acme.server =
"https://acme-staging-v02.api.letsencrypt.org/directory".
''
)
(mkRemovedOptionModule [ "security" "acme" "directory" ] "ACME Directory is now hardcoded to /var/lib/acme and its permisisons are managed by systemd. See https://github.com/NixOS/nixpkgs/issues/53852 for more info.")
(mkRemovedOptionModule [ "security" "acme" "preDelay" ] "This option has been removed. If you want to make sure that something executes before certificates are provisioned, add a RequiredBy=acme-\${cert}.service to the service you want to execute before the cert renewal")
(mkRemovedOptionModule [ "security" "acme" "activationDelay" ] "This option has been removed. If you want to make sure that something executes before certificates are provisioned, add a RequiredBy=acme-\${cert}.service to the service you want to execute before the cert renewal")
(mkChangedOptionModule [ "security" "acme" "validMin" ] [ "security" "acme" "validMinDays" ] (config: config.security.acme.validMin / (24 * 3600)))
];
2015-12-12 16:06:24 +01:00
config = mkMerge [
(mkIf (cfg.certs != { }) {
# FIXME Most of these custom warnings and filters for security.acme.certs.* are required
# because using mkRemovedOptionModule/mkChangedOptionModule with attrsets isn't possible.
warnings = filter (w: w != "") (mapAttrsToList (cert: data: if data.extraDomains != "_mkMergedOptionModule" then ''
The option definition `security.acme.certs.${cert}.extraDomains` has changed
to `security.acme.certs.${cert}.extraDomainNames` and is now a list of strings.
Setting a custom webroot for extra domains is not possible, instead use separate certs.
'' else "") cfg.certs);
assertions = let
certs = attrValues cfg.certs;
in [
{
assertion = cfg.email != null || all (certOpts: certOpts.email != null) certs;
message = ''
You must define `security.acme.certs.<name>.email` or
`security.acme.email` to register with the CA. Note that using
many different addresses for certs may trigger account rate limits.
'';
}
{
assertion = cfg.acceptTerms;
message = ''
You must accept the CA's terms of service before using
the ACME module by setting `security.acme.acceptTerms`
to `true`. For Let's Encrypt's ToS see https://letsencrypt.org/repository/
'';
}
] ++ (builtins.concatLists (mapAttrsToList (cert: data: [
{
assertion = data.user == "_mkRemovedOptionModule";
message = ''
The option definition `security.acme.certs.${cert}.user' no longer has any effect; Please remove it.
Certificate user is now hard coded to the "acme" user. If you would
like another user to have access, consider adding them to the
"acme" group or changing security.acme.certs.${cert}.group.
'';
}
{
assertion = data.allowKeysForGroup == "_mkRemovedOptionModule";
message = ''
The option definition `security.acme.certs.${cert}.allowKeysForGroup' no longer has any effect; Please remove it.
All certs are readable by the configured group. If this is undesired,
consider changing security.acme.certs.${cert}.group to an unused group.
'';
}
# * in the cert value breaks building of systemd services, and makes
# referencing them as a user quite weird too. Best practice is to use
# the domain option.
{
assertion = ! hasInfix "*" cert;
message = ''
The cert option path `security.acme.certs.${cert}.dnsProvider`
cannot contain a * character.
Instead, set `security.acme.certs.${cert}.domain = "${cert}";`
and remove the wildcard from the path.
'';
}
{
assertion = data.dnsProvider == null || data.webroot == null;
message = ''
Options `security.acme.certs.${cert}.dnsProvider` and
`security.acme.certs.${cert}.webroot` are mutually exclusive.
'';
}
]) cfg.certs));
2015-12-12 16:06:24 +01:00
users.users.acme = {
home = "/var/lib/acme";
group = "acme";
isSystemUser = true;
};
users.groups.acme = {};
systemd.services = {
"acme-fixperms" = userMigrationService;
} // (mapAttrs' (cert: conf: nameValuePair "acme-${cert}" conf.renewService) certConfigs)
// (optionalAttrs (cfg.preliminarySelfsigned) ({
"acme-selfsigned-ca" = selfsignCAService;
} // (mapAttrs' (cert: conf: nameValuePair "acme-selfsigned-${cert}" conf.selfsignService) certConfigs)));
systemd.timers = mapAttrs' (cert: conf: nameValuePair "acme-${cert}" conf.renewTimer) certConfigs;
systemd.targets = let
# Create some targets which can be depended on to be "active" after cert renewals
finishedTargets = mapAttrs' (cert: conf: nameValuePair "acme-finished-${cert}" {
wantedBy = [ "default.target" ];
requires = [ "acme-${cert}.service" ] ++ conf.selfsignedDeps;
after = [ "acme-${cert}.service" ] ++ conf.selfsignedDeps;
}) certConfigs;
# Create targets to limit the number of simultaneous account creations
# How it works:
# - Pick a "leader" cert service, which will be in charge of creating the account,
# and run first (requires + after)
# - Make all other cert services sharing the same account wait for the leader to
# finish before starting (requiredBy + before).
# Using a target here is fine - account creation is a one time event. Even if
# systemd clean --what=state is used to delete the account, so long as the user
# then runs one of the cert services, there won't be any issues.
accountTargets = mapAttrs' (hash: confs: let
leader = "acme-${(builtins.head confs).cert}.service";
dependantServices = map (conf: "acme-${conf.cert}.service") (builtins.tail confs);
in nameValuePair "acme-account-${hash}" {
requiredBy = dependantServices;
before = dependantServices;
requires = [ leader ];
after = [ leader ];
}) (groupBy (conf: conf.accountHash) (attrValues certConfigs));
in finishedTargets // accountTargets;
})
2015-12-12 16:06:24 +01:00
];
2016-05-09 07:53:27 +02:00
meta = {
maintainers = lib.teams.acme.members;
2016-05-09 07:53:27 +02:00
doc = ./acme.xml;
};
}