mailauth 7.2.0 → 7.3.1

package/lib/dkim2/fields.js

@@ -0,0 +1,571 @@
+'use strict';
+
+// Parsing, formatting and canonicalization of the two DKIM2 header fields, Message-Instance
+// and DKIM2-Signature (draft-ietf-dkim-dkim2-spec-06 sections 6, 7, 8 and 9.6)
+
+const { Buffer } = require('node:buffer');
+const { formatRelaxedLine, toLineBytes, isValidDnsName, isSameOrSubdomain } = require('../tools');
+const { trimChars, trimEndChars } = require('../trim');
+const { parseRecipe } = require('./recipe');
+
+const MI_KEY = 'message-instance';
+const SIG_KEY = 'dkim2-signature';
+
+// Section 4.1: header fields that are not signed, processed as if they were not present
+const UNSIGNED_FIELDS = new Set([
+    'apparently-to',
+    'arc-authentication-results',
+    'arc-message-signature',
+    'arc-seal',
+    'authentication-results',
+    'auto-submitted',
+    'delivered-to',
+    'dkim-signature',
+    'dl-expansion-history',
+    'original-recipient',
+    'received',
+    'return-path',
+    'sio-label-history',
+    'vbr-info',
+    'x400-received',
+    'x400-trace'
+]);
+
+const isUnsignedField = key => UNSIGNED_FIELDS.has(key) || key.startsWith('received-') || key.startsWith('x-');
+
+// Section 6.2: the header hash leaves out the unsigned fields and the DKIM2 fields themselves
+const isHashedField = key => !!key && key !== MI_KEY && key !== SIG_KEY && !isUnsignedField(key);
+
+const HASH_ALGORITHMS = new Set(['sha256', 'sha512']);
+
+// Section 8.3: at most 64 printable ASCII characters, no semicolon
+const NONCE = /^[\x21-\x3a\x3c-\x7e]{0,64}$/;
+// Section 2.14 textstring, without the surrounding FWS
+const TEXTSTRING = /^[A-Za-z0-9_-]+$/;
+// RFC 5234 1*DIGIT. t= values up to 10^12 have to be accepted (section 8.4), 15 digits keep every
+// value a safe integer
+const NUMBER = /^[0-9]{1,15}$/;
+// Section 2.14 base64string with the FWS removed. It "MUST be padded", so the length is a multiple of 4
+const BASE64 = /^(?:[A-Za-z0-9+/]{4})*(?:[A-Za-z0-9+/]{2}==|[A-Za-z0-9+/]{3}=)?$/;
+// RFC 5321 Domain, also used for selectors (section 3.5). The underscore is accepted as it is
+// in DKIM1 key record names
+const DOMAIN_LABEL = '[A-Za-z0-9_](?:[A-Za-z0-9_-]{0,61}[A-Za-z0-9_])?';
+const DOMAIN = new RegExp(`^${DOMAIN_LABEL}(?:\\.${DOMAIN_LABEL})*$`);
+// Section 7 and 8 x-tag-name
+const TAG_NAME = /^[A-Za-z][A-Za-z0-9_]*$/;
+
+const WSP_ALL = /[ \t\r\n]+/g;
+
+const isBase64 = value => !!value && BASE64.test(value);
+const isDomain = value => !!value && DOMAIN.test(value) && isValidDnsName(value);
+
+const toNumber = value => (NUMBER.test(value) ? Number(value) : NaN);
+
+/**
+ * Splits the value of a DKIM2 header field into its tags. Tag names are case insensitive
+ * (sections 7 and 8), so they are returned in lower case. Semicolons never occur inside a value.
+ *
+ * @param {String} value Header field value as a 'binary' string
+ * @returns {Object} { tags: Map of name to value with the FWS around it removed, error }
+ */
+const parseTags = value => {
+    let tags = new Map();
+    let segments = value.split(';');
+
+    for (let i = 0; i < segments.length; i++) {
+        let segment = trimChars(segments[i], ' \t\r\n');
+        if (!segment) {
+            // the tag-list ends every tag with a semicolon, and a final one can be left out
+            if (i < segments.length - 1) {
+                return { tags, error: 'empty tag' };
+            }
+            continue;
+        }
+
+        let eqPos = segment.indexOf('=');
+        if (eqPos < 0) {
+            return { tags, error: 'tag without a value' };
+        }
+
+        let name = trimEndChars(segment.substring(0, eqPos), ' \t\r\n');
+        if (!TAG_NAME.test(name)) {
+            return { tags, error: 'invalid tag name' };
+        }
+        name = name.toLowerCase();
+
+        if (tags.has(name)) {
+            return { tags, error: `duplicate tag ${name}` };
+        }
+
+        tags.set(name, segment.substring(eqPos + 1).replace(/^[ \t\r\n]+/, ''));
+    }
+
+    return { tags, error: null };
+};
+
+// the value of a header line, after the colon, as a 'binary' string
+const getFieldValue = line => {
+    let str = toLineBytes(line).toString('binary');
+    let colonPos = str.indexOf(':');
+    return colonPos >= 0 ? str.substring(colonPos + 1) : '';
+};
+
+// A list of base64strings, each of which can contain FWS (section 2.14)
+const splitBase64List = value => value.replace(WSP_ALL, '').split(',');
+
+// The value of a tag that can not contain whitespace, such as a number or a domain
+const plainValue = value => (/[ \t\r\n]/.test(value) ? null : value);
+
+/**
+ * Parses a Message-Instance header field (section 7)
+ *
+ * @param {Buffer|String} line Header line
+ * @returns {Object} { m, hashes: [{ algorithm, headerHash, bodyHash }], recipe, error }. `m` is
+ *          NaN when the m= tag could not be read, `recipe` is undefined without an r= tag
+ */
+const parseMessageInstance = line => {
+    let result = { m: NaN, hashes: [], recipe: undefined, error: null };
+    let { tags, error } = parseTags(getFieldValue(line));
+
+    let mValue = tags.get('m');
+    if (typeof mValue === 'string') {
+        result.m = toNumber(plainValue(mValue) || '');
+    }
+
+    const fail = message => {
+        result.error = message;
+        return result;
+    };
+
+    if (error) {
+        return fail('syntax error');
+    }
+
+    for (let tag of ['m', 'h']) {
+        if (!tags.has(tag)) {
+            return fail(`tag=${tag} missing`);
+        }
+    }
+
+    if (!(result.m >= 1)) {
+        return fail('syntax error');
+    }
+
+    // mi-h-tag: hash-set *("," hash-set), hash-set = hash-name ":" header-hash ":" body-hash
+    let seen = new Set();
+    for (let hashSet of tags.get('h').replace(WSP_ALL, '').split(',')) {
+        let parts = hashSet.split(':');
+        if (parts.length !== 3 || !TEXTSTRING.test(parts[0]) || !isBase64(parts[1]) || !isBase64(parts[2])) {
+            return fail('syntax error');
+        }
+
+        let algorithm = parts[0].toLowerCase();
+        if (seen.has(algorithm)) {
+            return fail('has a duplicate hash algorithm');
+        }
+        seen.add(algorithm);
+
+        result.hashes.push({ algorithm, headerHash: parts[1], bodyHash: parts[2] });
+    }
+
+    if (tags.has('r')) {
+        let recipeValue = tags.get('r').replace(WSP_ALL, '');
+        if (!isBase64(recipeValue)) {
+            return fail('syntax error');
+        }
+
+        let recipe = parseRecipe(Buffer.from(recipeValue, 'base64').toString('utf8'));
+        if (recipe.error) {
+            return fail(`contains invalid JSON: ${recipe.error}`);
+        }
+        result.recipe = recipe.value;
+    }
+
+    return result;
+};
+
+// Decodes an mf= or rt= path: the base64 encoded RFC 5321 reverse-path or Forward-path, angle
+// brackets included (sections 8.5 and 8.6). Returns false for a value that is not a path
+const decodePath = (value, allowNull) => {
+    if (!isBase64(value)) {
+        return false;
+    }
+    let path = Buffer.from(value, 'base64').toString('utf8');
+    if (!/^<[^<>\r\n]*>$/.test(path) || (path === '<>' && !allowNull)) {
+        return false;
+    }
+    return path;
+};
+
+/**
+ * Parses a DKIM2-Signature header field (section 8)
+ *
+ * @param {Buffer|String} line Header line
+ * @returns {Object} { i, m, t, nonce, mailFrom, rcptTo, nextDomain, signingDomain, signatures,
+ *          flags, error }. `i` is NaN when the i= tag could not be read
+ */
+const parseSignature = line => {
+    let result = {
+        i: NaN,
+        m: NaN,
+        t: NaN,
+        nonce: undefined,
+        mailFrom: undefined,
+        rcptTo: undefined,
+        nextDomain: undefined,
+        signingDomain: undefined,
+        signatures: [],
+        flags: [],
+        error: null
+    };
+
+    let { tags, error } = parseTags(getFieldValue(line));
+
+    let iValue = tags.get('i');
+    if (typeof iValue === 'string') {
+        result.i = toNumber(plainValue(iValue) || '');
+    }
+
+    const fail = message => {
+        result.error = message;
+        return result;
+    };
+
+    if (error) {
+        return fail('syntax error');
+    }
+
+    for (let tag of ['i', 'm', 't', 'd', 's']) {
+        if (!tags.has(tag)) {
+            return fail(`tag=${tag} missing`);
+        }
+    }
+
+    // there is either an nd= tag or both mf= and rt=, never both kinds
+    if (tags.has('nd')) {
+        for (let tag of ['mf', 'rt']) {
+            if (tags.has(tag)) {
+                return fail(`tag=${tag} was unexpected`);
+            }
+        }
+    } else {
+        for (let tag of ['mf', 'rt']) {
+            if (!tags.has(tag)) {
+                return fail(`tag=${tag} missing`);
+            }
+        }
+    }
+
+    result.m = toNumber(plainValue(tags.get('m')) || '');
+    result.t = toNumber(plainValue(tags.get('t')) || '');
+    if (!(result.i >= 1) || !(result.m >= 1) || isNaN(result.t)) {
+        return fail('syntax error');
+    }
+
+    result.signingDomain = plainValue(tags.get('d'));
+    if (!isDomain(result.signingDomain)) {
+        return fail('syntax error');
+    }
+    result.signingDomain = result.signingDomain.toLowerCase();
+
+    if (tags.has('nd')) {
+        result.nextDomain = plainValue(tags.get('nd'));
+        if (!isDomain(result.nextDomain)) {
+            return fail('syntax error');
+        }
+        result.nextDomain = result.nextDomain.toLowerCase();
+    } else {
+        result.mailFrom = decodePath(tags.get('mf').replace(WSP_ALL, ''), true);
+        result.rcptTo = splitBase64List(tags.get('rt')).map(value => decodePath(value, false));
+        if (!result.mailFrom || result.rcptTo.some(path => !path)) {
+            return fail('syntax error');
+        }
+    }
+
+    if (tags.has('n')) {
+        // section 8.3: implementations SHOULD reject a nonce longer than 64 characters
+        result.nonce = tags.get('n');
+        if (!NONCE.test(result.nonce)) {
+            return fail('syntax error');
+        }
+    }
+
+    if (tags.has('f')) {
+        for (let flag of tags.get('f').split(',')) {
+            flag = trimChars(flag, ' \t\r\n');
+            if (!TEXTSTRING.test(flag)) {
+                return fail('syntax error');
+            }
+            // flag values are case significant, and unknown ones are ignored
+            result.flags.push(flag);
+        }
+    }
+
+    // sig-s-tag: sig-set *("," sig-set), sig-set = selector ":" sig-name ":" message-sig
+    let selectors = new Set();
+    let algorithmCounts = new Map();
+    for (let sigSet of tags.get('s').replace(WSP_ALL, '').split(',')) {
+        let parts = sigSet.split(':');
+        if (parts.length !== 3 || !isDomain(parts[0]) || !TEXTSTRING.test(parts[1]) || !isBase64(parts[2])) {
+            return fail('syntax error');
+        }
+
+        let selector = parts[0].toLowerCase();
+        let algorithm = parts[1].toLowerCase();
+
+        if (selectors.has(selector)) {
+            return fail('has a duplicate selector');
+        }
+        selectors.add(selector);
+
+        // section 8.9: one additional signature with the same algorithm is allowed
+        let count = (algorithmCounts.get(algorithm) || 0) + 1;
+        if (count > 2) {
+            return fail('has more selectors than allowed');
+        }
+        algorithmCounts.set(algorithm, count);
+
+        result.signatures.push({ selector, algorithm, signature: parts[2] });
+    }
+
+    return result;
+};
+
+/**
+ * Canonicalizes a header field for the header hash (section 6.2). This is the same
+ * transformation as the DKIM1 "relaxed" header canonicalization
+ *
+ * @param {Buffer|String} line Header line
+ * @returns {Buffer} Canonicalized line, ending with CRLF
+ */
+const canonicalHeaderField = line => formatRelaxedLine(line, '\r\n');
+
+// Empties the signature values of the s= tag of a DKIM2-Signature value that has no whitespace
+// left in it, for the signature input (sections 9.6 and 11.6). Semicolons can not occur inside a
+// value, so splitting on them finds the tag the same way the parser does
+const blankSignatureValues = value =>
+    value
+        .split(';')
+        .map(segment => {
+            let eqPos = segment.indexOf('=');
+            if (eqPos < 0 || segment.substring(0, eqPos).toLowerCase() !== 's') {
+                return segment;
+            }
+            let sigSets = segment
+                .substring(eqPos + 1)
+                .split(',')
+                .map(sigSet => {
+                    // selector:sig-name:value, only the value is emptied. Anything else is a syntax
+                    // error that is reported anyway, and is left as it is
+                    let parts = sigSet.split(':');
+                    return parts.length === 3 ? `${parts[0]}:${parts[1]}:` : sigSet;
+                });
+            return `${segment.substring(0, eqPos + 1)}${sigSets.join(',')}`;
+        })
+        .join(';');
+
+/**
+ * Canonicalizes a Message-Instance or DKIM2-Signature header field for the signature input
+ * (section 9.6): the name in lower case, the field unfolded and every WSP character removed
+ *
+ * @param {Buffer|String} line Header line
+ * @param {Boolean} [blank] Empty the signature values of the s= tag
+ * @returns {Buffer} Canonicalized line, ending with CRLF
+ */
+const canonicalSignatureField = (line, blank) => {
+    let str = toLineBytes(line).toString('binary').replace(WSP_ALL, '');
+    let colonPos = str.indexOf(':');
+    let name = str.substring(0, colonPos).toLowerCase();
+    let value = str.substring(colonPos + 1);
+    if (blank) {
+        value = blankSignatureValues(value);
+    }
+    return Buffer.from(`${name}:${value}\r\n`, 'binary');
+};
+
+// The longest a folded line gets, not counting the CRLF
+const LINE_LENGTH = 76;
+const BASE64_CHUNK = 64;
+
+// Splits a base64 value into pieces a fold can go between. FWS is allowed anywhere inside a
+// base64string (section 2.14) and is ignored when the value is used. `lead` is the text in front
+// of the value, such as "s=selector:rsa-sha256:", and `space` tells if it starts a new tag
+const base64Atoms = (lead, value, trail, space) => {
+    let atoms = lead ? [{ text: lead, space: space !== false }] : [];
+    for (let pos = 0; pos < value.length; pos += BASE64_CHUNK) {
+        atoms.push({ text: value.substring(pos, pos + BASE64_CHUNK), space: false });
+    }
+    if (!atoms.length) {
+        atoms.push({ text: '', space: space !== false });
+    }
+    atoms[atoms.length - 1].text += trail || '';
+    return atoms;
+};
+
+// Joins the atoms of a header field, folding between them when a line would get too long. An
+// atom with `space` set is a new tag and gets a space in front of it when it is not folded
+const foldAtoms = (name, atoms) => {
+    let lines = [];
+    let line = `${name}:`;
+    for (let atom of atoms) {
+        let separator = atom.space ? ' ' : '';
+        if (line.length + separator.length + atom.text.length > LINE_LENGTH && line.length > name.length + 1) {
+            lines.push(line);
+            line = ` ${atom.text}`;
+        } else {
+            line += separator + atom.text;
+        }
+    }
+    lines.push(line);
+    return lines.join('\r\n');
+};
+
+const plainAtom = text => ({ text, space: true });
+
+/**
+ * Formats a Message-Instance header field
+ *
+ * @param {Object} options
+ * @param {Number} options.m Revision number
+ * @param {Object[]} options.hashes [{ algorithm, headerHash, bodyHash }]
+ * @param {Object} [options.recipe] Recipe object for the r= tag
+ * @returns {String} Folded header line, without the final CRLF
+ */
+const formatMessageInstance = ({ m, hashes, recipe }) => {
+    let atoms = [plainAtom(`m=${m};`)];
+
+    hashes.forEach((hash, index) => {
+        // a fold can go before the "," as FWS at the end of the previous base64string
+        atoms.push(
+            ...base64Atoms(`${index ? ',' : 'h='}${hash.algorithm}:`, hash.headerHash, ':', !index),
+            ...base64Atoms('', hash.bodyHash, index === hashes.length - 1 ? ';' : '')
+        );
+    });
+
+    if (recipe) {
+        atoms.push(...base64Atoms('r=', Buffer.from(JSON.stringify(recipe)).toString('base64'), ';'));
+    }
+
+    return foldAtoms('Message-Instance', atoms);
+};
+
+const encodePath = path => Buffer.from(path).toString('base64');
+
+/**
+ * Formats a DKIM2-Signature header field. Signature values that are not known yet are written
+ * as empty strings, which is the form the signature input uses (section 9.6)
+ *
+ * @param {Object} options
+ * @returns {String} Folded header line, without the final CRLF
+ */
+const formatSignature = ({ i, m, t, mailFrom, rcptTo, nextDomain, signingDomain, nonce, flags, signatures }) => {
+    let atoms = [plainAtom(`i=${i};`), plainAtom(`m=${m};`), plainAtom(`t=${t};`), plainAtom(`d=${signingDomain};`)];
+
+    if (nextDomain) {
+        atoms.push(plainAtom(`nd=${nextDomain};`));
+    } else {
+        atoms.push(...base64Atoms('mf=', encodePath(mailFrom), ';'));
+        rcptTo.forEach((path, index) => {
+            atoms.push(...base64Atoms(index ? ',' : 'rt=', encodePath(path), index === rcptTo.length - 1 ? ';' : '', !index));
+        });
+    }
+
+    if (typeof nonce === 'string') {
+        atoms.push(plainAtom(`n=${nonce};`));
+    }
+
+    if (flags?.length) {
+        atoms.push(plainAtom(`f=${flags.join(',')};`));
+    }
+
+    signatures.forEach((entry, index) => {
+        atoms.push(
+            ...base64Atoms(
+                `${index ? ',' : 's='}${entry.selector}:${entry.algorithm}:`,
+                entry.signature || '',
+                index === signatures.length - 1 ? ';' : '',
+                !index
+            )
+        );
+    });
+
+    return foldAtoms('DKIM2-Signature', atoms);
+};
+
+/**
+ * Splits an RFC 5321 path into its address, with the domain in lower case as section 11.4
+ * compares it. A source route in front of the mailbox is left out
+ *
+ * @param {String} path Path with or without the angle brackets
+ * @returns {Object} { address, domain }, both empty for the null path
+ */
+const parsePath = path => {
+    let address = (path || '').toString().trim().replace(/^<|>$/g, '');
+    if (address.charAt(0) === '@' && address.indexOf(':') > 0) {
+        address = address.substring(address.indexOf(':') + 1);
+    }
+    let atPos = address.lastIndexOf('@');
+    if (atPos < 0) {
+        return { address, domain: '' };
+    }
+    let domain = address.substring(atPos + 1).toLowerCase();
+    return { address: `${address.substring(0, atPos)}@${domain}`, domain };
+};
+
+/**
+ * Section 9.4: whether a hop continues the chain of custody of the hop before it. Its MAIL FROM
+ * domain has to match a RCPT TO of the previous hop after removing labels from the left, which is
+ * the same as being that domain or one of its subdomains. A hop with nd= has no MAIL FROM, and the
+ * null MAIL FROM has no domain, so the signing domain is matched instead: section 8.8 only waives
+ * the match of d= and mf= for the null MAIL FROM, never the chain of custody. A previous hop with
+ * nd= is checked against the d= of the next hop instead
+ *
+ * @param {Object} previous Parsed previous DKIM2-Signature
+ * @param {String} [mailFrom] MAIL FROM path of the hop
+ * @param {String} signingDomain Signing domain of the hop
+ * @returns {Boolean}
+ */
+const continuesCustody = (previous, mailFrom, signingDomain) => {
+    if (!previous || previous.nextDomain) {
+        return true;
+    }
+    let domain = mailFrom && mailFrom !== '<>' ? parsePath(mailFrom).domain : signingDomain;
+    return previous.rcptTo.some(path => isSameOrSubdomain(domain, parsePath(path).domain));
+};
+
+/**
+ * Builds the signature input of section 9.6: the Message-Instance fields by m=, the
+ * DKIM2-Signature fields before the signature by i=, and the signature itself with its signature
+ * values emptied
+ *
+ * @param {Array} instanceLines Message-Instance header lines in m= order
+ * @param {Array} signatureLines Earlier DKIM2-Signature header lines in i= order
+ * @param {Buffer|String} ownLine The DKIM2-Signature that is signed or verified
+ * @returns {Buffer}
+ */
+const buildSignatureInput = (instanceLines, signatureLines, ownLine) =>
+    Buffer.concat([
+        ...instanceLines.map(line => canonicalSignatureField(line)),
+        ...signatureLines.map(line => canonicalSignatureField(line)),
+        canonicalSignatureField(ownLine, true)
+    ]);
+
+module.exports = {
+    MI_KEY,
+    SIG_KEY,
+    HASH_ALGORITHMS,
+    NONCE,
+    TEXTSTRING,
+    isHashedField,
+    isDomain,
+    parseMessageInstance,
+    parseSignature,
+    canonicalHeaderField,
+    canonicalSignatureField,
+    formatMessageInstance,
+    formatSignature,
+    parsePath,
+    isBase64,
+    continuesCustody,
+    buildSignatureInput
+};

package/lib/dkim2/hash.d.ts

@@ -0,0 +1,3 @@
+// Type definitions for mailauth/lib/dkim2/hash
+
+export { dkim2Hash } from '../../index';

package/lib/dkim2/hash.js

@@ -0,0 +1,51 @@
+'use strict';
+
+const { writeToStream } = require('../tools');
+const { Dkim2Parser } = require('./dkim2-parser');
+const { HASH_ALGORITHMS } = require('./fields');
+
+// Reads a message and keeps its DKIM2 state: the hashed header fields and the body hashes
+class HashParser extends Dkim2Parser {
+    constructor(algorithms) {
+        super();
+        algorithms.forEach(algorithm => this.hashBodyWith(algorithm));
+    }
+
+    async finalChunk() {
+        this.state = this.currentState();
+    }
+}
+
+/**
+ * Computes the DKIM2 header and body hashes of a message as they go into the h= tag of a
+ * Message-Instance header field (draft-ietf-dkim-dkim2-spec-06 sections 6 and 7.3)
+ *
+ * @param {ReadableStream|Buffer|String} input RFC 5322 message
+ * @param {Object} [options]
+ * @param {String[]} [options.algorithms=['sha256']] "sha256" and/or "sha512"
+ * @returns {Object} { headers, hashes }. `headers` lists the names of the hashed header fields in
+ *          the order they are hashed, `hashes` is [{ algorithm, headerHash, bodyHash }]
+ * @throws {Error} For an unsupported hash algorithm
+ */
+const dkim2Hash = async (input, options) => {
+    let algorithms = Array.from(new Set((options?.algorithms || ['sha256']).map(algorithm => String(algorithm).toLowerCase())));
+    for (let algorithm of algorithms) {
+        if (!HASH_ALGORITHMS.has(algorithm)) {
+            let err = new Error(`Unsupported hash algorithm ${algorithm}, use sha256 or sha512`);
+            err.code = 'EINVALIDALGO';
+            throw err;
+        }
+    }
+
+    let parser = new HashParser(algorithms);
+    await writeToStream(parser, input);
+    let { state } = parser;
+
+    return {
+        // section 6.2: sorted by name
+        headers: Array.from(state.headers.keys()).sort(),
+        hashes: algorithms.map(algorithm => ({ algorithm, headerHash: state.headerHash(algorithm), bodyHash: state.bodyHash(algorithm) }))
+    };
+};
+
+module.exports = { dkim2Hash };