security: triage the CodeQL backlog — 129 alerts, three of them live (#561)

* fix(web): escape quotes in every HTML escaper, not just & < > The escapers are all `div.textContent = x; return div.innerHTML`. That round-trip escapes &, < and > -- the only characters the HTML serializer must escape in a text node -- and leaves quotes alone. Every widget then interpolates the result into a quoted attribute value: value="${escapeHtml(v)}" title="${escapeHtml(v)}" so a value of `x" onmouseover="alert(1)` closes the attribute and adds an event handler of its own. CodeQL reported this 83 times (js/incomplete-html-attribute-sanitization) across the widget files. It is one bug, not 83: the widgets each carry a standalone fallback that did escape quotes, but they all prefer BaseWidget.escapeHtml when window.BaseWidget exists -- which it always does in the shipped page -- so the correct fallbacks were dead code and the incomplete shared one ran. Fixed at each source instead of at the call sites. app-shell.js already documented this exact gap in a comment and worked around it by building DOM nodes by hand; that workaround stays (setting a property cannot be got wrong), the comment is now accurate. cache.html's delete button interpolated the cache key into `onclick="deleteCacheFile('...')"`. Escaping cannot help there -- the browser HTML-decodes the attribute before parsing it as JS, so `'` becomes a real `'` again -- so the key moves to a data-cache-key attribute that the handler reads back. url-input.js additionally wrote a value straight into an <a href> after validating it against a schema-supplied protocol list, and that list accepted any RFC 3986 scheme -- "javascript" included. Scriptable schemes (javascript, data, vbscript, blob, filesystem) are now refused both when the list is normalised and when a URL is checked against it, and the render path routes its href through the same check instead of emitting whatever was stored (js/xss-through-dom). test/js/unit/test_html_escaping.js reads each escaper out of the shipped file and runs it, so losing the quote handling again fails a test rather than a scan. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(security): stop request-supplied names from reaching paths outside their base Three of the py/path-injection alerts were live, not lint: * GET /api/v3/plugins/<plugin_id>/static/<path:file_path> read any file whose resolved path *string-prefixed* the plugin directory. Flask's default converter forbids a slash but not dots, and get_plugin_directory('..') returned the parent of the plugins directory because it exists -- so every file under the project root then prefixed that directory, config/config_secrets.json included. The prefix check was also wrong on its own terms: with plugin dir "plugin-repos/foo", "../foo-evil/x" resolves to "plugin-repos/foo-evil/x", whose string does start with "plugin-repos/foo". * POST /api/v3/plugins/of-the-day/json/delete interpolated the request body's file_id into f"{file_id}.json" and unlinked it, unvalidated. A file_id of "../../../../etc/something" deleted that file. This is the one finding in the batch that destroyed data rather than exposing it. * POST /api/v3/cache/delete passed the body's key through CacheManager.clear_cache to DiskCache, which joined it as a filename and called os.remove. Same shape, same result. The guard goes in DiskCache.get_cache_path, the single choke point get/set/clear share, so every caller is covered rather than just this route. Real keys are the stems of files already flat in the cache directory -- that is how list_cache_files derives them -- so nothing legitimate is turned away. The rest of the cluster (web_interface/app.py's asset route, the plugin update handler, _get_plugin_version, the plugin-schema read in config.py) was guarded in ways that held, but each had grown its own version of the check. They now go through one helper, src/common/path_safety.py, which returns the *sanitised value* rather than a verdict -- so a caller cannot validate one string and open another, which is how the two real bugs above were shaped. Also: WiFiManager.connect_to_network took the SSID and password straight from POST /api/v3/wifi/connect into nmcli's argv. There is no shell there, so CodeQL's py/command-line-injection alert overstates the risk -- but nmcli reads a leading "-" as an option, so an SSID of "--ask" asks nmcli to run differently rather than to join a network. Both values are now checked for shape (802.11's 32-octet SSID limit, WPA's 8-63 char passphrase or 64-char hex key, no control characters, no leading dash) before any subprocess runs. test/test_path_traversal_guards.py asserts on the filesystem, not just the status code: a handler that returns 403 and deletes the file anyway would pass the weaker check. Twelve of its cases fail against the unpatched code. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(web): refuse a plugin id that is not a plain name, don't truncate it pages_v3 and scripts/dev_server.py ran request ids through os.path.basename and carried on with what came out, so "../weather" rendered the config form for "weather". Nothing escaped the plugins directory -- the relative_to guards held -- but the handler answered a request nobody made, and validating one string while the filesystem sees another is the shape both live traversals earlier in this branch had. Same treatment as the rest: safe_path_component rejects rather than truncates, resolve_under returns the path it checked, and the call sites use what those return. The three handlers that had hand-rolled resolve-and-relative_to blocks lose about twenty lines to the shared one. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * docs(web): say what the plugin web_ui iframe actually is The docstring claimed the fragment runs "in a sandboxed iframe". The iframe in plugin_config.html carries no sandbox attribute, so the fragment runs with the interface's own origin. That is fine -- the file belongs to an installed plugin, and an installed plugin already runs Python on the device, so the trust boundary is install rather than this route -- but a comment promising containment that is not there is worse than no comment. This is the context for the py/reflective-xss alert on this handler. Also drops the now-unused os/os.path imports. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(web): inline url-input's scheme guard at the previewLink.href sink CodeQL flagged this line as a new high-severity js/xss-through-dom alert on this PR even though it is already covered by SCRIPTABLE_SCHEMES: the guard reached the sink through safeHref -> isValidUrl, two function calls away, which its DOM-based-XSS sanitizer recognition does not trace. Behavior is unchanged -- same scheme check, same SCRIPTABLE_SCHEMES list, same allowedProtocols gate -- just inlined directly above the previewLink.href assignment it guards, so the barrier is visible in the same scope as the sink. Added a regression test that runs the shipped onInput handler (not just the extracted helpers) against a mocked DOM, so a future change that reintroduces an unguarded previewLink.href assignment fails here. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(security): address CodeRabbit findings on the CodeQL triage PR - src/wifi_manager.py: reject non-ASCII WPA-PSK passphrases before any credential-saving or connect flow runs. NetworkManager only accepts printable ASCII passphrases (or a 64-char hex key); a non-ASCII value was previously saved/attempted before nmcli itself rejected it. - web_interface/blueprints/api_v3/config.py: fail closed when the plugin config schema path can't be resolved under the plugins directory (e.g. a symlinked plugin dir). Previously this fell through with secret_fields left empty, so submitted credentials for that plugin were saved as ordinary, unencrypted configuration. - web_interface/static/v3/js/widgets/plugin-file-manager.js: stop splicing the JSON day/column key into an inline oninput="..." handler string. escHtml() escapes quotes for a normal HTML attribute, but the browser HTML-decodes the attribute before running it as script, which undoes that escaping and lets a crafted column name (e.g. from an uploaded JSON file) break out of the JS string and execute. Cell edits now travel through data-day/data-col attributes read by one delegated 'input' listener instead. While in this file: fixed 6 pre-existing missing-')' typos on multi-line safeSetHTML(...) calls (already flagged by Biome in this PR's own CodeRabbit run as syntax errors blocking its lint pass). These predate this PR (present on main too) but made the whole file fail to parse in any JS engine, which is a bigger problem than the XSS finding itself and directly touches the same lines. Added/extended regression tests for each fix; full suites pass (pytest: 4580 passed, 62 skipped; JS: 84 assertions). Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
2026-10-04 14:25:08 +00:00 · 2026-09-11 15:32:03 -04:00
parent 5137e86d16
commit f6367d63ae
27 changed files with 1520 additions and 170 deletions
@@ -0,0 +1,320 @@
+// Every HTML escaper in the web interface must escape quotes, not just
+// & < >.
+//
+// The escapers are all written as `div.textContent = x; return div.innerHTML`.
+// That round-trip escapes &, < and > because those are the only characters the
+// HTML serializer has to escape in a *text node* -- quotes are left alone. But
+// the widgets interpolate the result into quoted attribute values
+// (`value="${escapeHtml(v)}"`, `title="${escapeHtml(v)}"`, ...), and there a
+// bare `"` closes the attribute and lets the value add attributes of its own:
+//
+//     name" onfocus="alert(1)
+//
+// CodeQL reported 83 js/incomplete-html-attribute-sanitization alerts for
+// exactly this. This suite pins the fix at the source: it reads each real
+// implementation out of the shipped file and runs it, so an escaper that loses
+// its quote handling again fails here rather than in a scanner run weeks later.
+
+const fs = require('fs');
+const path = require('path');
+
+const ROOT = path.resolve(__dirname, '../../../web_interface');
+
+let pass = 0, fail = 0;
+const ok = (label, cond, extra) => cond
+  ? (pass++, console.log('  ok   ' + label))
+  : (fail++, console.log('  FAIL ' + label + (extra !== undefined ? '  -> ' + JSON.stringify(extra).slice(0, 300) : '')));
+
+// ── DOM shim ───────────────────────────────────────────────────────────────
+// Mirrors what a browser does reading innerHTML back off textContent: & < >
+// are escaped, quotes are not. Anything the escaper adds on top of that is
+// the escaper's own doing, which is what we are testing.
+class FakeEl {
+  set textContent(v) { this._t = String(v == null ? '' : v); }
+  get textContent() { return this._t || ''; }
+  get innerHTML() {
+    return (this._t || '')
+      .replace(/&/g, '&amp;')
+      .replace(/</g, '&lt;')
+      .replace(/>/g, '&gt;');
+  }
+}
+global.document = { createElement: () => new FakeEl() };
+global.window = global;
+
+// ── source extraction ──────────────────────────────────────────────────────
+// Pull a function out of a real source file by its opening line and balanced
+// braces, so the test runs the shipped code rather than a copy of it.
+function extract(file, opener) {
+  const src = fs.readFileSync(path.join(ROOT, file), 'utf8');
+  const start = src.indexOf(opener);
+  if (start < 0) {
+    console.error(`FAIL: cannot find ${JSON.stringify(opener)} in ${file}`);
+    process.exit(1);
+  }
+  let i = src.indexOf('{', start), depth = 0;
+  for (let j = i; j < src.length; j++) {
+    if (src[j] === '{') depth++;
+    else if (src[j] === '}') {
+      depth--;
+      if (depth === 0) return src.slice(start, j + 1);
+    }
+  }
+  console.error(`FAIL: unbalanced braces after ${JSON.stringify(opener)} in ${file}`);
+  process.exit(1);
+}
+
+// Evaluate an extracted escaper and return it as a callable.
+function loadFn(file, opener, name, { method = false } = {}) {
+  const body = extract(file, opener);
+  // Class/object methods (`escapeHtml(text) {...}`) are not valid statements on
+  // their own -- wrap them in an object literal so they can be evaluated.
+  const code = method
+    ? `(function(){ const o = { ${body} }; return o.${name}.bind(o); })()`
+    : `(function(){ ${body}; return ${name}; })()`;
+  // eslint-disable-next-line no-eval
+  return eval(code);
+}
+
+// ── the escapers, as shipped ───────────────────────────────────────────────
+const ESCAPERS = [
+  ['base-widget.js (BaseWidget.escapeHtml)',
+   'static/v3/js/widgets/base-widget.js', 'escapeHtml(text) {', 'escapeHtml', true],
+  ['plugins_manager.js (top-level escapeHtml)',
+   'static/v3/plugins_manager.js', 'function escapeHtml(text) {', 'escapeHtml', false],
+  ['plugins_manager.js (starlark escapeHtml)',
+   'static/v3/plugins_manager.js', 'function escapeHtml(str) {', 'escapeHtml', false],
+  ['error_handler.js (escapeHtml)',
+   'static/v3/js/utils/error_handler.js', 'function escapeHtml(text) {', 'escapeHtml', false],
+  ['json-file-manager.js (_esc)',
+   'static/v3/js/widgets/json-file-manager.js', '_esc(str) {', '_esc', true],
+  ['plugin-file-manager.js (escHtml)',
+   'static/v3/js/widgets/plugin-file-manager.js', 'function escHtml(s) {', 'escHtml', false],
+  ['app-shell.js (escapeHtml)',
+   'static/v3/js/app-shell.js', 'escapeHtml(text) {', 'escapeHtml', true],
+  ['logs.html (escapeHtml)',
+   'templates/v3/partials/logs.html', 'function escapeHtml(text) {', 'escapeHtml', false],
+  ['cache.html (escapeHtml)',
+   'templates/v3/partials/cache.html', 'function escapeHtml(text) {', 'escapeHtml', false],
+];
+
+// The breakout payload: closes a double-quoted attribute and opens an event
+// handler. If `"` survives escaping, this is live script in the rendered page.
+const BREAKOUT = 'x" onmouseover="alert(1)';
+
+console.log('\n1. every escaper neutralises a double-quote attribute breakout');
+for (const [label, file, opener, name, method] of ESCAPERS) {
+  const fn = loadFn(file, opener, name, { method });
+  const out = String(fn(BREAKOUT));
+  ok(`${label}: no raw "`, !out.includes('"'), out);
+  ok(`${label}: emits &quot;`, out.includes('&quot;'), out);
+}
+
+console.log('\n2. every escaper also escapes single quotes');
+for (const [label, file, opener, name, method] of ESCAPERS) {
+  const fn = loadFn(file, opener, name, { method });
+  const out = String(fn("x' onmouseover='alert(1)"));
+  ok(`${label}: no raw '`, !out.includes("'"), out);
+}
+
+console.log('\n3. the & < > behaviour they already had is unchanged');
+for (const [label, file, opener, name, method] of ESCAPERS) {
+  const fn = loadFn(file, opener, name, { method });
+  const out = String(fn('<img src=x onerror=alert(1)> & done'));
+  ok(`${label}: no raw <`, !out.includes('<'), out);
+  ok(`${label}: no raw >`, !out.includes('>'), out);
+  ok(`${label}: & becomes &amp;`, /&amp;/.test(out), out);
+}
+
+console.log('\n4. ampersands are escaped before quotes, so &quot; cannot be forged');
+// If `&` were escaped last, the input `&quot;` would come out as a literal
+// `"` after the browser decodes the attribute. Order matters; pin it.
+for (const [label, file, opener, name, method] of ESCAPERS) {
+  const fn = loadFn(file, opener, name, { method });
+  const out = String(fn('&quot;'));
+  ok(`${label}: &quot; input stays inert`, out === '&amp;quot;', out);
+}
+
+// ── url-input scheme handling (js/xss-through-dom) ─────────────────────────
+console.log('\n5. url-input never treats a scriptable scheme as a valid URL');
+{
+  const src = fs.readFileSync(path.join(ROOT, 'static/v3/js/widgets/url-input.js'), 'utf8');
+  const a = src.indexOf('const RFC_SCHEME_PATTERN');
+  const b = src.indexOf("window.LEDMatrixWidgets.register('url-input'");
+  if (a < 0 || b < 0) { console.error('FAIL: cannot locate url-input helpers'); process.exit(1); }
+  // eslint-disable-next-line no-eval
+  const helpers = eval(`(function(){ ${src.slice(a, b)}; return { normalizeProtocols, isValidUrl, safeHref }; })()`);
+  const { normalizeProtocols, isValidUrl, safeHref } = helpers;
+
+  ok('javascript: rejected under the default protocols',
+     !isValidUrl('javascript:alert(1)', ['http', 'https']));
+  ok('javascript: still rejected when a schema asks for it',
+     !isValidUrl('javascript:alert(1)', normalizeProtocols(['javascript'])));
+  ok('a schema asking only for javascript falls back to http/https',
+     JSON.stringify(normalizeProtocols(['javascript'])) === JSON.stringify(['http', 'https']),
+     normalizeProtocols(['javascript']));
+  ok('data: rejected', !isValidUrl('data:text/html,<script>alert(1)</script>', ['http', 'https', 'data']));
+  ok('vbscript: rejected', !isValidUrl('vbscript:msgbox(1)', ['http', 'https', 'vbscript']));
+  ok('normalizeProtocols drops scriptable schemes but keeps the rest',
+     JSON.stringify(normalizeProtocols('https,javascript,ftp')) === JSON.stringify(['https', 'ftp']),
+     normalizeProtocols('https,javascript,ftp'));
+
+  ok('ordinary https URL still valid', isValidUrl('https://example.com/x?y=1', ['http', 'https']));
+  ok('ordinary http URL still valid', isValidUrl('http://example.com', ['http', 'https']));
+  ok('a scheme outside the allow-list is still rejected',
+     !isValidUrl('ftp://example.com', ['http', 'https']));
+
+  ok('safeHref passes a good URL through', safeHref('https://example.com', ['http', 'https']) === 'https://example.com');
+  ok('safeHref blanks a javascript: URL', safeHref('javascript:alert(1)', ['http', 'https']) === '');
+  ok('safeHref blanks an unparseable value', safeHref('not a url', ['http', 'https']) === '');
+}
+
+// ── url-input onInput: the sink itself, not just the pulled-out helpers ────
+// The scheme check now lives inline in onInput, right where the value reaches
+// `previewLink.href`, instead of behind safeHref/isValidUrl -- see the comment
+// at that assignment in url-input.js for why. Run the handler as shipped so a
+// regression that reintroduces an unguarded `previewLink.href = value` fails
+// here, not just in a scanner run weeks later.
+console.log('\n6. url-input onInput: previewLink.href is guarded at the sink');
+{
+  class FakeClassList {
+    constructor() { this.classes = new Set(['hidden']); }
+    add(c) { this.classes.add(c); }
+    remove(c) { this.classes.delete(c); }
+    contains(c) { return this.classes.has(c); }
+  }
+  const mockDoc = (value, protocolsAttr) => {
+    const elements = {
+      _input: { value, checkValidity: () => true, validationMessage: '', classList: new FakeClassList() },
+      _preview: { classList: new FakeClassList() },
+      _preview_link: {
+        classList: new FakeClassList(),
+        _href: undefined,
+        set href(v) { this._href = v; },
+        get href() { return this._href; },
+        removeAttribute(name) { if (name === 'href') this._href = undefined; },
+      },
+      _widget: { dataset: { protocols: protocolsAttr } },
+      _error: { classList: new FakeClassList(), textContent: '' },
+    };
+    return {
+      elements,
+      getElementById: (id) => elements[Object.keys(elements).find(k => id === `field${k}`)] || null,
+    };
+  };
+
+  function runOnInput(value, protocolsAttr) {
+    const { elements, getElementById } = mockDoc(value, protocolsAttr);
+    const savedDocument = global.document;
+    const savedWindow = global.window;
+    let registered = null;
+    global.document = { createElement: () => new FakeEl(), getElementById };
+    global.window = {
+      LEDMatrixWidgets: {
+        register: (name, obj) => { registered = obj; },
+        get: () => registered,
+        getHandlers: () => registered.handlers,
+      },
+    };
+    try {
+      const src = fs.readFileSync(path.join(ROOT, 'static/v3/js/widgets/url-input.js'), 'utf8');
+      // eslint-disable-next-line no-eval
+      eval(src);
+      registered.handlers.onInput('field');
+    } finally {
+      global.document = savedDocument;
+      global.window = savedWindow;
+    }
+    return elements;
+  }
+
+  let els = runOnInput('javascript:alert(1)', 'http,https');
+  ok('javascript: never reaches previewLink.href', els._preview_link.href === undefined, els._preview_link.href);
+  ok('javascript: leaves the preview hidden', els._preview.classList.contains('hidden'));
+
+  els = runOnInput('https://example.com', 'http,https');
+  ok('an ordinary https URL reaches previewLink.href', els._preview_link.href === 'https://example.com', els._preview_link.href);
+  ok('an ordinary https URL unhides the preview', !els._preview.classList.contains('hidden'));
+
+  els = runOnInput('data:text/html,<script>alert(1)</script>', 'http,https,data');
+  ok('data: never reaches previewLink.href even when the schema allows it',
+     els._preview_link.href === undefined, els._preview_link.href);
+
+  els = runOnInput('not a url', 'http,https');
+  ok('an unparseable value never reaches previewLink.href', els._preview_link.href === undefined, els._preview_link.href);
+}
+
+// ── plugin-file-manager: cell edits travel via data-*, not inline handlers ──
+// A JSON key/day from an uploaded file used to be spliced, HTML-escaped,
+// into an oninput="...('${escHtml(col)}'...)" attribute. escHtml neutralises
+// a quote for an ordinary attribute, but here the value also has to survive
+// as a *JS string literal* -- the browser HTML-decodes the attribute before
+// running it as script, which turns the escaped quote back into a real one
+// and lets a column named `x');alert(1);//` break out of the string and run
+// arbitrary JS. Cell edits now reach _pfmCellEdit only via data-day/data-col
+// read by one delegated listener, so this pins that no inline handler string
+// is built from the value at all.
+console.log("\n7. plugin-file-manager: cell edits never go through an inline handler string");
+{
+  const src = fs.readFileSync(path.join(ROOT, 'static/v3/js/widgets/plugin-file-manager.js'), 'utf8');
+
+  function extractFn(opener) {
+    const start = src.indexOf(opener);
+    if (start < 0) { console.error(`FAIL: cannot find ${JSON.stringify(opener)}`); process.exit(1); }
+    let i = src.indexOf('{', start), depth = 0;
+    for (let j = i; j < src.length; j++) {
+      if (src[j] === '{') depth++;
+      else if (src[j] === '}') { depth--; if (depth === 0) return src.slice(start, j + 1); }
+    }
+    console.error(`FAIL: unbalanced braces after ${JSON.stringify(opener)}`);
+    process.exit(1);
+  }
+
+  const escHtmlFn = loadFn('static/v3/js/widgets/plugin-file-manager.js', 'function escHtml(s) {', 'escHtml', false);
+  const renderEntryTableSrc = extractFn('function renderEntryTable(fieldId, container, content) {');
+
+  const calls = [];
+  const fakeWindow = { _pfmCellEdit: (fieldId, day, col, value) => calls.push({ fieldId, day, col, value }) };
+
+  class FakeContainer {
+    constructor() { this._html = ''; this._listeners = {}; }
+    set innerHTML(v) { this._html = v; }
+    get innerHTML() { return this._html; }
+    set textContent(v) { this._html = v; }
+    addEventListener(type, fn) { this._listeners[type] = fn; }
+    dispatch(type, target) { this._listeners[type]({ target }); }
+  }
+
+  function fakeCell(day, col, value) {
+    return {
+      closest: (sel) => (sel.includes('data-day') ? { dataset: { day: String(day), col: String(col) }, value } : null),
+    };
+  }
+
+  // eslint-disable-next-line no-eval
+  const renderEntryTable = eval(`(function(getState, escHtml, safeSetHTML, window){
+    ${renderEntryTableSrc}
+    return renderEntryTable;
+  })`)(
+    () => ({ entriesPerPage: 20, _tablePage: 1 }),
+    escHtmlFn,
+    (target, html) => { target.innerHTML = html; },
+    fakeWindow
+  );
+
+  const maliciousCol = "x');alert(1);//";
+  const container = new FakeContainer();
+  renderEntryTable('field1', container, { '1': { [maliciousCol]: 'hello' } });
+
+  ok('no inline oninput handler is emitted for a cell', !/oninput=/.test(container.innerHTML), container.innerHTML);
+  ok('the malicious column name never appears unescaped in the markup',
+     !container.innerHTML.includes(maliciousCol), container.innerHTML);
+
+  container.dispatch('input', fakeCell('1', maliciousCol, 'typed value'));
+  ok('the delegated listener still reaches _pfmCellEdit with the real day/col',
+     calls.length === 1 && calls[0].day === '1' && calls[0].col === maliciousCol && calls[0].value === 'typed value',
+     calls);
+}
+
+console.log(`\n${pass} passed, ${fail} failed\n`);
+process.exit(fail ? 1 : 0);