mox was already strict in its "\r\n.\r\n" handling for end-of-message in an
smtp transaction.
due to a mostly unrelated bug, sequences of "\nX\n", including "\n.\n" were
rejected with a "local processing error".
the sequence "\r\n.\n" dropped the dot, not necessarily a big problem, this is
unlikely to happen in a legimate transaction and the behaviour not
unreasonable.
we take this opportunity to reject all bare \r. we detect all slightly
incorrect combinations of "\r\n.\r\n" with an error mentioning smtp smuggling,
in part to appease the tools checking for it.
smtp errors are 500 "bad syntax", and mention smtp smuggling.
to get the security benefits (detecting mitm attempts), explicitly configure
clients to use a scram plus variant, e.g. scram-sha-256-plus. unfortunately,
not many clients support it yet.
imapserver scram plus support seems to work with the latest imtest (imap test
client) from cyrus-sasl. no success yet with mutt (with gsasl) though.
so a single user cannot fill up the disk.
by default, there is (still) no limit. a default can be set in the config file
for all accounts, and a per-account max size can be set that would override any
global setting.
this does not take into account disk usage of the index database. and also not
of any file system overhead.
when writing the 4xx temporary error line, we were taking 1s in between each
byte. the total line could take longer than 30 seconds, which is the timeout we
use for reading a whole line (regardless of individual bytes). so mox as
deliverer was timing out to mox as slow rejecter. this causes slow writes to
not take longer than the 30s timeout: if we are 2s before the 30s, we write the
remainder in one go.
based on a debug log from naturalethic, thanks!
- prometheus is now behind an interface, they aren't dependencies for the
reusable components anymore.
- some dependencies have been inverted: instead of packages importing a main
package to get configuration, the main package now sets configuration in
these packages. that means fewer internals are pulled in.
- some functions now have new parameters for values that were retrieved from
package "mox-".
we don't want external software to include internal details like mlog.
slog.Logger is/will be the standard.
we still have mlog for its helper functions, and its handler that logs in
concise logfmt used by mox.
packages that are not meant for reuse still pass around mlog.Log for
convenience.
we use golang.org/x/exp/slog because we also support the previous Go toolchain
version. with the next Go release, we'll switch to the builtin slog.
we already allowed it for (authenticated) SMTP submission. it turns out also
legitimate senders can use this invalid syntax to deliver messages.
for issue #101 by Fell, thanks for reporting & explaining!
both cases are quite typical for spammers, and not for legitimate senders.
this doesn't apply to known senders. and it only requires that the content look
more like ham instead of spam. so legitimate mail can still get through with
these properties.
so facilitate debugging. a remote client that logs details about failing
connections can give the cid to the mox operator to find the relevant logging.
- dmarc reports: add a cid to the log line about one run of sending reports, and log line for each report
- in smtpclient, also handle tls errors from the first read after a handshake. we appear to sometimes get tls alerts about bad certificates on the first read.
- for messages to dmarc/tls reporting addresses that we think should/can not be processed as reports, add an X-Mox- header explaining the reason.
- tls reports: send report messages with From address of postmaster at an actually configured domain for the mail host. and only send reports when dkim signing is configured for that domain. the domain is also the submitter domain. the rfc seems to require dkim-signing with an exact match with the message from and submitter.
- for incoming tls reports, in the smtp server, we do allow a dkim-signature domain that is higher-level (up to publicsuffix) of the message from domain. so we are stricter in what we send than what we receive.
we were already accepting, processing and displaying incoming tls reports. now
we start tracking TLS connection and security-policy-related errors for
outgoing message deliveries as well. we send reports once a day, to the
reporting addresses specified in TLSRPT records (rua) of a policy domain. these
reports are about MTA-STS policies and/or DANE policies, and about
STARTTLS-related failures.
sending reports is enabled by default, but can be disabled through setting
NoOutgoingTLSReports in mox.conf.
only at the end of the implementation process came the realization that the
TLSRPT policy domain for DANE (MX) hosts are separate from the TLSRPT policy
for the recipient domain, and that MTA-STS and DANE TLS/policy results are
typically delivered in separate reports. so MX hosts need their own TLSRPT
policies.
config for the per-host TLSRPT policy should be added to mox.conf for existing
installs, in field HostTLSRPT. it is automatically configured by quickstart for
new installs. with a HostTLSRPT config, the "dns records" and "dns check" admin
pages now suggest the per-host TLSRPT record. by creating that record, you're
requesting TLS reports about your MX host.
gathering all the TLS/policy results is somewhat tricky. the tentacles go
throughout the code. the positive result is that the TLS/policy-related code
had to be cleaned up a bit. for example, the smtpclient TLS modes now reflect
reality better, with independent settings about whether PKIX and/or DANE
verification has to be done, and/or whether verification errors have to be
ignored (e.g. for tls-required: no header). also, cached mtasts policies of
mode "none" are now cleaned up once the MTA-STS DNS record goes away.
firstly by using crypto/rand in those cases. and secondly by putting a lock
around the Read (though it isn't used at the moment).
found while working while implementing sending tls reports.
before this change, a message in the rejects folder that was read and marked as
notjunk (e.g. automatically by webmail), could cause a dmarc report to be sent
for another junky message from the domain. we now require positive signals to
be for messages not in the rejects mailbox.
the text/plain body of a dmarc report contains the period, but it was in local
time while claiming to be in utc. make it utc, so we often get nicely rounded
whole 24h utc days.
- more eagerly report about overrides, so domain owners can better tell that
switching from p=none to p=reject will not cause trouble for these messages.
- report multiple reasons, e.g. mailing list and sampled out
- in dmarc analysis for rejects from first-time senders (possibly spammers),
fix the conditional check on nonjunk messages.
- in evaluations view in admin, show unaligned spf pass in yellow too and a few
more small tweaks.
"fatal" was meant as "we need fatal for the connection, it will be dropped".
but it sounds more serious, as if something needs to be fixed.
hopefully enough for issue #39 by ArnoSen
in smtpserver, we store dmarc evaluations (under the right conditions).
in dmarcdb, we periodically (hourly) send dmarc reports if there are
evaluations. for failed deliveries, we deliver the dsn quietly to a submailbox
of the postmaster mailbox.
this is on by default, but can be disabled in mox.conf.
with requiretls, the tls verification mode/rules for email deliveries can be
changed by the sender/submitter. in two ways:
1. "requiretls" smtp extension to always enforce verified tls (with mta-sts or
dnssec+dane), along the entire delivery path until delivery into the final
destination mailbox (so entire transport is verified-tls-protected).
2. "tls-required: no" message header, to ignore any tls and tls verification
errors even if the recipient domain has a policy that requires tls verification
(mta-sts and/or dnssec+dane), allowing delivery of non-sensitive messages in
case of misconfiguration/interoperability issues (at least useful for sending
tls reports).
we enable requiretls by default (only when tls is active), for smtp and
submission. it can be disabled through the config.
for each delivery attempt, we now store (per recipient domain, in the account
of the sender) whether the smtp server supports starttls and requiretls. this
support is shown (after having sent a first message) in the webmail when
sending a message (the previous 3 bars under the address input field are now 5
bars, the first for starttls support, the last for requiretls support). when
all recipient domains for a message are known to implement requiretls,
requiretls is automatically selected for sending (instead of "default" tls
behaviour). users can also select the "fallback to insecure" to add the
"tls-required: no" header.
new metrics are added for insight into requiretls errors and (some, not yet
all) cases where tls-required-no ignored a tls/verification error.
the admin can change the requiretls status for messages in the queue. so with
default delivery attempts, when verified tls is required by failing, an admin
could potentially change the field to "tls-required: no"-behaviour.
messages received (over smtp) with the requiretls option, get a comment added
to their Received header line, just before "id", after "with".
getting mox to compile required changing code in only a few places where
package "syscall" was used: for accessing file access times and for umask
handling. an open problem is how to start a process as an unprivileged user on
windows. that's why "mox serve" isn't implemented yet. and just finding a way
to implement it now may not be good enough in the near future: we may want to
starting using a more complete privilege separation approach, with a process
handling sensitive tasks (handling private keys, authentication), where we may
want to pass file descriptors between processes. how would that work on
windows?
anyway, getting mox to compile for windows doesn't mean it works properly on
windows. the largest issue: mox would normally open a file, rename or remove
it, and finally close it. this happens during message delivery. that doesn't
work on windows, the rename/remove would fail because the file is still open.
so this commit swaps many "remove" and "close" calls. renames are a longer
story: message delivery had two ways to deliver: with "consuming" the
(temporary) message file (which would rename it to its final destination), and
without consuming (by hardlinking the file, falling back to copying). the last
delivery to a recipient of a message (and the only one in the common case of a
single recipient) would consume the message, and the earlier recipients would
not. during delivery, the already open message file was used, to parse the
message. we still want to use that open message file, and the caller now stays
responsible for closing it, but we no longer try to rename (consume) the file.
we always hardlink (or copy) during delivery (this works on windows), and the
caller is responsible for closing and removing (in that order) the original
temporary file. this does cost one syscall more. but it makes the delivery code
(responsibilities) a bit simpler.
there is one more obvious issue: the file system path separator. mox already
used the "filepath" package to join paths in many places, but not everywhere.
and it still used strings with slashes for local file access. with this commit,
the code now uses filepath.FromSlash for path strings with slashes, uses
"filepath" in a few more places where it previously didn't. also switches from
"filepath" to regular "path" package when handling mailbox names in a few
places, because those always use forward slashes, regardless of local file
system conventions. windows can handle forward slashes when opening files, so
test code that passes path strings with forward slashes straight to go stdlib
file i/o functions are left unchanged to reduce code churn. the regular
non-test code, or test code that uses path strings in places other than
standard i/o functions, does have the paths converted for consistent paths
(otherwise we would end up with paths with mixed forward/backward slashes in
log messages).
windows cannot dup a listening socket. for "mox localserve", it isn't
important, and we can work around the issue. the current approach for "mox
serve" (forking a process and passing file descriptors of listening sockets on
"privileged" ports) won't work on windows. perhaps it isn't needed on windows,
and any user can listen on "privileged" ports? that would be welcome.
on windows, os.Open cannot open a directory, so we cannot call Sync on it after
message delivery. a cursory internet search indicates that directories cannot
be synced on windows. the story is probably much more nuanced than that, with
long deep technical details/discussions/disagreement/confusion, like on unix.
for "mox localserve" we can get away with making syncdir a no-op.
the vendored dns resolver code is a copy of the go stdlib dns resolver, with
awareness of the "authentic data" (i.e. dnssec secure) added, as well as support
for enhanced dns errors, and looking up tlsa records (for dane). ideally it
would be upstreamed, but the chances seem slim.
dnssec-awareness is added to all packages, e.g. spf, dkim, dmarc, iprev. their
dnssec status is added to the Received message headers for incoming email.
but the main reason to add dnssec was for implementing dane. with dane, the
verification of tls certificates can be done through certificates/public keys
published in dns (in the tlsa records). this only makes sense (is trustworthy)
if those dns records can be verified to be authentic.
mox now applies dane to delivering messages over smtp. mox already implemented
mta-sts for webpki/pkix-verification of certificates against the (large) pool
of CA's, and still enforces those policies when present. but it now also checks
for dane records, and will verify those if present. if dane and mta-sts are
both absent, the regular opportunistic tls with starttls is still done. and the
fallback to plaintext is also still done.
mox also makes it easy to setup dane for incoming deliveries, so other servers
can deliver with dane tls certificate verification. the quickstart now
generates private keys that are used when requesting certificates with acme.
the private keys are pre-generated because they must be static and known during
setup, because their public keys must be published in tlsa records in dns.
autocert would generate private keys on its own, so had to be forked to add the
option to provide the private key when requesting a new certificate. hopefully
upstream will accept the change and we can drop the fork.
with this change, using the quickstart to setup a new mox instance, the checks
at internet.nl result in a 100% score, provided the domain is dnssec-signed and
the network doesn't have any issues.
increase() and rate() don't seem to assume a previous value of 0 when a vector
gets a first value for a label. you would think that an increase() on a
first-value mox_panic_total{"..."}=1 would return 1, and similar for rate(), but
that doesn't appear to be the behaviour. so we just explicitly initialize the
count to 0 for each possible label value. mox has more vector metrics, but
panics feels like the most important, and it's too much code to initialize them
all, for all combinations of label values. there is probably a better way that
fixes this for all cases...
we match messages to their parents based on the "references" and "in-reply-to"
headers (requiring the same base subject), and in absense of those headers we
also by only base subject (against messages received max 4 weeks ago).
we store a threadid with messages. all messages in a thread have the same
threadid. messages also have a "thread parent ids", which holds all id's of
parent messages up to the thread root. then there is "thread missing link",
which is set when a referenced immediate parent wasn't found (but possibly
earlier ancestors can still be found and will be in thread parent ids".
threads can be muted: newly delivered messages are automatically marked as
read/seen. threads can be marked as collapsed: if set, the webmail collapses
the thread to a single item in the basic threading view (default is to expand
threads). the muted and collapsed fields are copied from their parent on
message delivery.
the threading is implemented in the webmail. the non-threading mode still works
as before. the new default threading mode "unread" automatically expands only
the threads with at least one unread (not seen) meessage. the basic threading
mode "on" expands all threads except when explicitly collapsed (as saved in the
thread collapsed field). new shortcuts for navigation/interaction threads have
been added, e.g. go to previous/next thread root, toggle collapse/expand of
thread (or double click), toggle mute of thread. some previous shortcuts have
changed, see the help for details.
the message threading are added with an explicit account upgrade step,
automatically started when an account is opened. the upgrade is done in the
background because it will take too long for large mailboxes to block account
operations. the upgrade takes two steps: 1. updating all message records in the
database to add a normalized message-id and thread base subject (with "re:",
"fwd:" and several other schemes stripped). 2. going through all messages in
the database again, reading the "references" and "in-reply-to" headers from
disk, and matching against their parents. this second step is also done at the
end of each import of mbox/maildir mailboxes. new deliveries are matched
immediately against other existing messages, currently no attempt is made to
rematch previously delivered messages (which could be useful for related
messages being delivered out of order).
the threading is not yet exposed over imap.
for submission, we don't care about the value. users typically won't be able to
easily fix the errors (of their mail client software). so we just ignore the
domain/ip address, unless in pedantic mode.
this also parses "additional information after literal addresses" more
strictly/correctly.
for issue #55 by gimpf, thanks for the report!
the recent webmail addition added localserve local delivery in queue.Add, so we
just that for smtpserver too.
and don't drop incoming smtp deliver messages, but deliver as normal.
without public ip's, the generated mox config will try to listen on 0.0.0.0 and
::, but because there is already a listener for 127.0.0.1:80 (and possibly
others), a bind for 0.0.0.0:80 will fail. explicit public ip's are needed.
the public http listener is useful for ACME validation over http.
for issue #52
this is based on @bobobo1618's PR #50. bobobo1618 had the right idea, i tried
including an "is forwarded email" configuration option but that indeed became
too tightly coupled. the "is forwarded" option is still planned, but it is
separate from the "accept rejects to mailbox" config option, because one could
still want to push back on forwarded spam messages.
we do an actual accept, delivering to a configured mailbox, instead of storing
to the rejects mailbox where messages can automatically be removed from. one
of the goals of mox is not pretend to accept email while actually junking it.
users can still configure delivery to a junk folder (as was already possible),
but aren't deleted automatically. there is still an X-Mox-Reason header in the
message, and a log line about accepting the reject, but otherwise it is
registered and treated as an (smtp) accept.
the ruleset mailbox is still required to keep that explicit. users can specify
Inbox again.
hope this is good enough for PR #50, otherwise we'll change it.
soon, we can have multiple rejects mailboxes. and checking against the
configured rejects mailbox name wasn't foolproof to begin with, because it may
have changed between delivery to the rejects mailbox and the message being
moved.
after upgrading, messages currently in rejects mailboxes don't have IsReject
set, so they don't get the special rejecs treatment when being moved. they are
removed from the rejects mailbox after some time though, and newly added
rejects will be treated correctly. so this means some existing messages wrongly
delivered to the rejects mailbox, and moved out, aren't used (for a positive
signal) for future deliveries. saves a bit of complexity in the
implementation. i think the tradeoff is worth it.
related to discussion in issue #50
it was far down on the roadmap, but implemented earlier, because it's
interesting, and to help prepare for a jmap implementation. for jmap we need to
implement more client-like functionality than with just imap. internal data
structures need to change. jmap has lots of other requirements, so it's already
a big project. by implementing a webmail now, some of the required data
structure changes become clear and can be made now, so the later jmap
implementation can do things similarly to the webmail code. the webmail
frontend and webmail are written together, making their interface/api much
smaller and simpler than jmap.
one of the internal changes is that we now keep track of per-mailbox
total/unread/unseen/deleted message counts and mailbox sizes. keeping this
data consistent after any change to the stored messages (through the code base)
is tricky, so mox now has a consistency check that verifies the counts are
correct, which runs only during tests, each time an internal account reference
is closed. we have a few more internal "changes" that are propagated for the
webmail frontend (that imap doesn't have a way to propagate on a connection),
like changes to the special-use flags on mailboxes, and used keywords in a
mailbox. more changes that will be required have revealed themselves while
implementing the webmail, and will be implemented next.
the webmail user interface is modeled after the mail clients i use or have
used: thunderbird, macos mail, mutt; and webmails i normally only use for
testing: gmail, proton, yahoo, outlook. a somewhat technical user is assumed,
but still the goal is to make this webmail client easy to use for everyone. the
user interface looks like most other mail clients: a list of mailboxes, a
search bar, a message list view, and message details. there is a top/bottom and
a left/right layout for the list/message view, default is automatic based on
screen size. the panes can be resized by the user. buttons for actions are just
text, not icons. clicking a button briefly shows the shortcut for the action in
the bottom right, helping with learning to operate quickly. any text that is
underdotted has a title attribute that causes more information to be displayed,
e.g. what a button does or a field is about. to highlight potential phishing
attempts, any text (anywhere in the webclient) that switches unicode "blocks"
(a rough approximation to (language) scripts) within a word is underlined
orange. multiple messages can be selected with familiar ui interaction:
clicking while holding control and/or shift keys. keyboard navigation works
with arrows/page up/down and home/end keys, and also with a few basic vi-like
keys for list/message navigation. we prefer showing the text instead of
html (with inlined images only) version of a message. html messages are shown
in an iframe served from an endpoint with CSP headers to prevent dangerous
resources (scripts, external images) from being loaded. the html is also
sanitized, with javascript removed. a user can choose to load external
resources (e.g. images for tracking purposes).
the frontend is just (strict) typescript, no external frameworks. all
incoming/outgoing data is typechecked, both the api request parameters and
response types, and the data coming in over SSE. the types and checking code
are generated with sherpats, which uses the api definitions generated by
sherpadoc based on the Go code. so types from the backend are automatically
propagated to the frontend. since there is no framework to automatically
propagate properties and rerender components, changes coming in over the SSE
connection are propagated explicitly with regular function calls. the ui is
separated into "views", each with a "root" dom element that is added to the
visible document. these views have additional functions for getting changes
propagated, often resulting in the view updating its (internal) ui state (dom).
we keep the frontend compilation simple, it's just a few typescript files that
get compiled (combined and types stripped) into a single js file, no additional
runtime code needed or complicated build processes used. the webmail is served
is served from a compressed, cachable html file that includes style and the
javascript, currently just over 225kb uncompressed, under 60kb compressed (not
minified, including comments). we include the generated js files in the
repository, to keep Go's easily buildable self-contained binaries.
authentication is basic http, as with the account and admin pages. most data
comes in over one long-term SSE connection to the backend. api requests signal
which mailbox/search/messages are requested over the SSE connection. fetching
individual messages, and making changes, are done through api calls. the
operations are similar to imap, so some code has been moved from package
imapserver to package store. the future jmap implementation will benefit from
these changes too. more functionality will probably be moved to the store
package in the future.
the quickstart enables webmail on the internal listener by default (for new
installs). users can enable it on the public listener if they want to. mox
localserve enables it too. to enable webmail on existing installs, add settings
like the following to the listeners in mox.conf, similar to AccountHTTP(S):
WebmailHTTP:
Enabled: true
WebmailHTTPS:
Enabled: true
special thanks to liesbeth, gerben, andrii for early user feedback.
there is plenty still to do, see the list at the top of webmail/webmail.ts.
feedback welcome as always.
the space is explicitly mentioned as not valid in rfc 5321, but it clients do
send it, such as microsoft outlook 365 apps for enterprise. no need to punish
such users, we'll allow it. but only for submission, not regular smtp, because
it is normally a sign of a spammer. we still don't allow it in pedantic mode
(as used by localserve).
for issue #51 by hmfaysal, thanks for reporting and testing!
for conditional storing and quick resynchronisation (not sure if mail clients are actually using it that).
each message now has a "modseq". it is increased for each change. with
condstore, imap clients can request changes since a certain modseq. that
already allows quickly finding changes since a previous connection. condstore
also allows storing (e.g. setting new message flags) only when the modseq of a
message hasn't changed.
qresync should make it fast for clients to get a full list of changed messages
for a mailbox, including removals.
we now also keep basic metadata of messages that have been removed (expunged).
just enough (uid, modseq) to tell client that the messages have been removed.
this does mean we have to be careful when querying messages from the database.
we must now often filter the expunged messages out.
we also keep "createseq", the modseq when a message was created. this will be
useful for the jmap implementation.
the warnings that remained were either unused code that i wanted to use in the
future, or other type's of todo's. i've been mentally ignoring them, assuming i
would get back to them soon enough to fix them. but that hasn't happened yet,
and it's better to have a clean list with only actual isses.
so mail user agents will show DSNs threaded/grouped with the original message.
we store the MessageID in the message queue, so we have the value within reach
when we need it.
i saw a references header in a DSN from gmail on a test account. makes sense to me.
an EHLO ipv4 address looks like this: "[1.2.3.4]". for ipv6, the syntax is:
"[IPv6🔡:1]". mail user agents aren't as careful in compliance as smtp
servers. for incoming messages from smtp servers, we want to be strict (we're
eager to find a reason not to accept spam messages, and not adhering to the
standards is usually a strong spam signal), but there is no reason to punish
authenticated users.
for the syntax requirements, see ABNF rule "address-literal" in rfc 5321.
for issue #48 by @bobobo1618, thanks!
we could make more types of delays configurable. the current approach isn't
great, as it results in an a default value of "0s" in the config file, while
the actual default is 15s (which is documented just above, but still).
