[pmg-devel] [RFC PATCH] fix-2971: DKIM: Add setting to use From header when signing
Stoiko Ivanov
s.ivanov at proxmox.com
Wed Jan 10 22:39:17 CET 2024
Thanks for tackling this!
Finally got around to look a bit into the RFCs...
(and currently don't think the premises of DMARC work too well with many
e-mail setups (not only lists))
the bugreport has gained quite a few comments and further suggestions that
we should consider:
* fixed signing domain for all mails (something the DKIM RFC orginally
envisioned)
* signing bounces to not have them generate DMARC reports (maybe even a
general fallback from envelope-from (RFC5321.From) to header-from
(RFC5322.From) if the former would not be signed, but the latter would?
* out-of-office mails
at least the first two might make sense to consider (out-of-office mails
usually are sent as bounce (empty 5321.from))
maybe the setting could be "dkim_prefer_rfc5322_domain" (better names more
than welcome), and the choice on, which domain to actually use for the
'd=' tag in the signature could be done in PMG::DKIMSign::signing_domain.
to sign bounces originating from PMG (e.g. the daily spam and admin
reports) - you'd need to add the DKIM-signing to
PMG::Utils::reinject_local_mail (and to test what happens with mails sent
via '/usr/sbin/sendmail' on the PMG (e.g. smartd-messages and
cronjob-outputs).
The changes do neither to happen with this patchset nor do you need to do
them, but maybe some of it offers itself for implementation, but please
mention 'partially fixes #2971' somewhere.
Some documentation for this would be needed as well.
further comments inline:
On Wed, 3 Jan 2024 15:53:29 +0100
Maximiliano Sandoval <m.sandoval at proxmox.com> wrote:
> We add an option to use the address from the `From:` header instead of
> the Envelope From address. Following RFC 5322 [2], we use the `Sender:`
> header when there are multiple addresses in the `From:` header.
sorry, I know I pointed you in that direction - but the DMARC rfc (where
this enhancement request initially originated) says - most of those cases
don't happen in DMARC protected domains (reality might have a different
view on that):
https://datatracker.ietf.org/doc/html/rfc7489#section-6.6.1
so we can probably spare ourselves the going through multiple headers,
and can take the part after the first '@' until either EOL or '>'
as domain (but please verify this with the acceptable syntax in rfc 5322)
>
> From RFC 6376 [1]:
wrong rfc - this is from the dmarc-one:
https://datatracker.ietf.org/doc/html/rfc7489#section-3.1.1
(DKIM itself doesn't require the signing domain to occur anywhere in the
SMTP-dialogue, or the mail-headers or body:
https://datatracker.ietf.org/doc/html/rfc6376#section-3.11 (INFORMATIVE DISCUSSION))
>
> To illustrate, in relaxed mode, if a validated DKIM signature
> successfully verifies with a "d=" domain of "example.com", and the
> RFC5322.From address is "alerts at news.example.com", the DKIM "d="
> domain and the RFC5322.From domain are considered to be "in
> alignment". In strict mode, this test would fail, since the "d="
> domain does not exactly match the FQDN of the address.
>
> Current Problems:
>
> - `dkim_sign_rfc5322` is probably not a descriptive name for the setting
> - The setting description might be improved upon
>
> Tested with the following command:
>
> swaks --from foo at bar1 --to EMAIL -s PMG_ADDR:26 --data "Date: %DATE%\nTo: %TO_ADDRESS%\nFrom: foo at bar2\nSubject: test %DATE%\nMessage-Id: <%MESSAGEID%>\nX-Mailer: swaks v%SWAKS_VERSION% jetmore.org/john/code/swaks/\n%NEW_HEADERS%\n%BODY%\n"
>
> [1] https://datatracker.ietf.org/doc/html/rfc6376
> [2] https://datatracker.ietf.org/doc/html/rfc5322#section-3.6.2
>
> Signed-off-by: Maximiliano Sandoval <m.sandoval at proxmox.com>
> ---
> src/PMG/Config.pm | 7 +++++++
> src/PMG/DKIMSign.pm | 43 +++++++++++++++++++++++++++++++++++++++-
> src/PMG/RuleDB/Accept.pm | 2 +-
> src/PMG/RuleDB/BCC.pm | 2 +-
> src/bin/pmg-smtp-filter | 1 +
> 5 files changed, 52 insertions(+), 3 deletions(-)
>
> diff --git a/src/PMG/Config.pm b/src/PMG/Config.pm
> index 7339e0d..b6c61e1 100644
> --- a/src/PMG/Config.pm
> +++ b/src/PMG/Config.pm
> @@ -134,6 +134,11 @@ EODESC
> description => "Default DKIM selector",
> type => 'string', format => 'dns-name', #see RFC6376 3.1
> },
> + dkim_sign_rfc5322 => {
> + description => "Whether to use the address from the From header when DKIM signing outbound mails.",
> + type => 'boolean',
> + default => 0,
> + },
> };
> }
>
> @@ -152,6 +157,7 @@ sub options {
> dkim_sign => { optional => 1 },
> dkim_sign_all_mail => { optional => 1 },
> dkim_selector => { optional => 1 },
> + dkim_sign_rfc5322 => { optional => 1 },
> };
> }
>
> @@ -1788,6 +1794,7 @@ my $pmg_service_params = {
> dkim_selector => 1,
> dkim_sign => 1,
> dkim_sign_all_mail => 1,
> + dkim_sign_rfc5322 => 1,
> },
> };
>
> diff --git a/src/PMG/DKIMSign.pm b/src/PMG/DKIMSign.pm
> index 08197f8..009313d 100644
> --- a/src/PMG/DKIMSign.pm
> +++ b/src/PMG/DKIMSign.pm
> @@ -84,8 +84,49 @@ sub signing_domain {
> }
>
>
> +sub parse_signing_sender {
> + # If there is exactly one single address in the 'From' header we use that
> + # address for signing, otherwise we use the 'Sender' header.
> + my ($entity) = @_;
> +
> + my $from_count = 0;
> + my @from_headers = $entity->head->get('from');
> + foreach my $from_header (@from_headers) {
> + my @senders = split(',', $from_header);
this won't work in a robust manner - consider the following (valid) from
header:
```
From: "Ivanov, Stoiko" <s.ivanov at proxmox.com>
```
> + $from_count += scalar(@senders);
> + }
> +
> + if ($from_count > 1) {
> + my $send_count = 0;
> + my @send_headers = $entity->head->get('sender');
> + foreach my $send_header (@send_headers) {
> + my @senders = split(',', $send_header);
> + $send_count += scalar(@senders);
> + }
> + if ($send_count > 1) {
> + syslog('warning', "Email has more than one address in its 'Sender' " .
> + "it won't be signed. See RFC 5322 Section 3.6.2");
> + } elsif ($send_count == 1) {
> + return $entity->head->get('sender', 0);
> + } else {
> + syslog('warning', "Email has multiple addresses in its 'From' header " .
> + "and no 'Sender' header, it won't be signed. See RFC 5322 Section 3.6.2");
> + }
> + } elsif ($from_count == 1) {
> + return $entity->head->get('from', 0);
> + }
> +}
> +
> +
> sub sign_entity {
> - my ($entity, $selector, $sender, $sign_all) = @_;
> + my ($entity, $dkim, $sender) = @_;
> +
> + my $sign_all = $dkim->{sign_all};
> + my $selector = $dkim->{selector};
> +
> + if ($dkim->{sign_rfc5322}) {
> + $sender = parse_signing_sender($entity);
> + }
>
> die "no selector provided\n" if ! $selector;
>
> diff --git a/src/PMG/RuleDB/Accept.pm b/src/PMG/RuleDB/Accept.pm
> index 4ebd6da..2e1bfa7 100644
> --- a/src/PMG/RuleDB/Accept.pm
> +++ b/src/PMG/RuleDB/Accept.pm
> @@ -103,7 +103,7 @@ sub execute {
> if ($dkim->{sign}) {
> eval {
> $entity = PMG::DKIMSign::sign_entity($entity,
> - $dkim->{selector}, $msginfo->{sender}, $dkim->{sign_all});
> + $dkim, $msginfo->{sender});
> };
> syslog('warning',
> "Could not create DKIM-Signature - disabling Signing: $@") if $@;
> diff --git a/src/PMG/RuleDB/BCC.pm b/src/PMG/RuleDB/BCC.pm
> index 0f016f8..6db48af 100644
> --- a/src/PMG/RuleDB/BCC.pm
> +++ b/src/PMG/RuleDB/BCC.pm
> @@ -143,7 +143,7 @@ sub execute {
> if ($dkim->{sign}) {
> eval {
> $entity = PMG::DKIMSign::sign_entity($entity,
> - $dkim->{selector}, $msginfo->{sender}, $dkim->{sign_all});
> + $dkim, $msginfo->{sender});
> };
> syslog('warning',
> "Could not create DKIM-Signature - disabling Signing: $@") if $@;
> diff --git a/src/bin/pmg-smtp-filter b/src/bin/pmg-smtp-filter
> index 7da3de8..549dbd4 100755
> --- a/src/bin/pmg-smtp-filter
> +++ b/src/bin/pmg-smtp-filter
> @@ -627,6 +627,7 @@ sub handle_smtp {
> my $dkim_sign = $msginfo->{trusted} && $pmg_cfg->get('admin', 'dkim_sign');
> if ($dkim_sign) {
> $msginfo->{dkim}->{sign} = $dkim_sign;
> + $msginfo->{dkim}->{sign_rfc5322} = $pmg_cfg->get('admin', 'dkim_sign_rfc5322');
> $msginfo->{dkim}->{sign_all} = $pmg_cfg->get('admin', 'dkim_sign_all_mail');
> $msginfo->{dkim}->{selector} = $pmg_cfg->get('admin', 'dkim_selector');
> }
More information about the pmg-devel
mailing list