harden: stop the background reading the shared state singleton, and enforce it at build time (closes #324)
All checks were successful
check / check (push) Successful in 33s
e2e / e2e-chrome (push) Successful in 1m45s
e2e / e2e-firefox (push) Successful in 31s

Five defects, one of which destroyed every wallet, came from src/background reading and writing the module-level state singleton the MV3 worker never populates, which silently served DEFAULT_STATE. Each point fix created the next defect. The background now has its own per-call getState() and a queued read-modify-write updateState(); the singleton is unreachable from it, and an unpopulated read throws instead of serving defaults.

The prohibition is enforced by the build, not by review: build.js asserts over esbuild's own metafile that no forbidden module is an input of a background bundle, so every specifier syntax esbuild resolves is covered, and both halves of the table are checked for rot -- a stale key, a stale module, an empty list, or an unlisted entry point under src/background/ all fail the build. The ESLint rule remains as fast local feedback and reads the same shared table. Known bounds are documented where the table lives.

Also closes #320: getProvider() now requires a validated network id, so a cold worker no longer prepares a non-mainnet dApp transaction for mainnet and gets refused by the wallet's own verifier. backgroundRefresh() no longer mutates address objects across a network round trip, the broadcast path takes its endpoint and chain id from one snapshot, and eight test storage stubs now structured-clone on get as the real chrome.storage.local does.

closes #320
This commit was merged in pull request #344.
This commit is contained in:
2026-08-23 17:57:30 +02:00
parent 36bc6bee0e
commit bd0a626e7b
40 changed files with 2959 additions and 653 deletions

View File

@@ -0,0 +1,206 @@
// ESLint rule: the background bundle may not reach the shared state singleton.
//
// src/shared/state.js holds a module-level `state` object, loaded once by
// loadState() and mutated in place from then on. That is the popup's model. In
// the MV3 service worker there is no "once": the worker is terminated when
// idle and revived by the next message, nothing loads state at module scope,
// and an unpopulated read used to be served DEFAULT_STATE without complaint —
// five defects, one cause
// (https://git.eeqj.de/sneak/AutistMask/issues/324). The background has its
// own per-call storage layer in src/background/state.js instead.
//
// THIS RULE IS NOT THE GUARANTEE, and must not be described as one. The
// guarantee is in build.js: FORBIDDEN_INPUTS / assertNoForbiddenInputs() fails
// the build when esbuild's own metafile reports src/shared/state.js as an input
// of a background bundle. That consults the resolution esbuild actually
// performed, so no specifier syntax and no resolution rule can slip past it,
// and Dockerfile:42 runs `make build` in CI.
//
// What this rule is: fast local feedback, in the editor and in `make lint`,
// before a full bundle. It reads sources from disk and matches import
// specifiers TEXTUALLY, so it is a best-effort approximation of module
// resolution — a hand-rolled matcher will diverge from a real bundler, and two
// earlier revisions of this file proved it by shipping holes (a template
// literal, a dynamic `import()`, a comment inside the call, a directory
// resolved through `package.json` `main`). Those are all covered now, and the
// next divergence is caught by the build rather than by widening this again.
//
// It checks REACHABILITY, not just the direct require: the singleton is one
// `require()` away from any shared module the background pulls in, and a
// re-export would put it back in the bundle without any background file naming
// it. So each background file is the root of a walk over the CommonJS require
// graph, and the error names the whole chain that brought the singleton in.
//
// Matching textually over-approximates — a specifier inside a comment or a
// string counts — which is the safe direction here: the failure mode is a
// spurious error naming an exact file and line, not a silent hole.
//
// Two shapes this rule does NOT report, both of which the build does fail on
// (each measured with `make lint` and `make build` on the branch that added
// this note):
//
// - a computed specifier, `require("../shared/" + "state")` — esbuild
// constant-folds it, so it is in the bundle and `make build` is exit 2
// naming src/shared/state.js, while `make lint` is exit 0. Same for
// `import("../shared/" + variable)`, which esbuild resolves as a glob.
// - a symlink to the module — esbuild reports the real path and fails the
// build; this rule resolves the link's own path and sees a different file.
//
// Both are pinned as non-reports in tests/backgroundStateLintRule.test.js, so
// this list is a measured description of the rule rather than a claim about
// it. They are known divergences, not things that cannot happen. A
// matcher will keep diverging from a bundler; that is why the guarantee is the
// build's and this rule is not widened again to chase them.
const fs = require("fs");
const path = require("path");
const { FORBIDDEN_INPUTS } = require("../forbiddenBundleInputs");
// The modules to keep out, repo-relative, taken from the same table build.js
// asserts against so that the two layers cannot name different paths. A second
// literal copy here is how a rename disarms one of them while the other still
// looks enforced.
const FORBIDDEN = [...new Set(Object.values(FORBIDDEN_INPUTS).flat())];
// Whatever may sit between a keyword, a paren and a specifier: whitespace and
// comments. `import(/* webpackChunkName: "x" */ "./x")` is a standard bundler
// idiom, and an inline `/* eslint-… */` is just as ordinary, so a matcher that
// allows only \s there is not strict, it is broken. Each alternative starts
// with a distinct character, so this cannot backtrack quadratically.
const GAP = "(?:\\s|/\\*[^]*?\\*/|//[^\\n]*)";
const SPECIFIER = "[\"'`]([^\"'`]+)[\"'`]";
// Both alternatives capture the specifier: call form first
// (`require(...)`/`import(...)`), then clause form (`from "x"`, and the bare
// side-effect `import "x"`). Nothing after the specifier is matched, so a
// trailing comment or a trailing comma cannot break the match either.
const SPECIFIER_RE = new RegExp(
`\\b(?:require|import)${GAP}*\\(${GAP}*${SPECIFIER}` +
`|\\b(?:from|import)${GAP}+${SPECIFIER}`,
"g",
);
// The `main` of a directory's package.json, as a specifier relative to that
// directory, or null. esbuild resolves a directory through it, so a walk that
// stops at `<dir>/index.js` reports a specifier it matched perfectly well as
// unresolvable.
function packageMain(dir) {
try {
const pkg = JSON.parse(
fs.readFileSync(path.join(dir, "package.json"), "utf8"),
);
return typeof pkg.main === "string" && pkg.main ? pkg.main : null;
} catch {
return null;
}
}
// Resolve a relative require to a file path, trying what node and esbuild would
// in the order they would: the path itself, then extensions, then the directory
// (its package.json `main`, then its index.js).
function resolveRelative(fromFile, spec) {
if (!spec.startsWith(".")) return null; // a package, not our tree
const base = path.resolve(path.dirname(fromFile), spec);
const main = packageMain(base);
for (const candidate of [
base,
base + ".js",
base + ".json",
...(main
? [path.resolve(base, main), path.resolve(base, main) + ".js"]
: []),
path.join(base, "index.js"),
]) {
try {
if (fs.statSync(candidate).isFile()) return candidate;
} catch {
// Not this candidate.
}
}
return null;
}
function requiresOf(file) {
let source;
try {
source = fs.readFileSync(file, "utf8");
} catch {
return [];
}
const out = [];
for (const match of source.matchAll(SPECIFIER_RE)) {
const resolved = resolveRelative(file, match[1] ?? match[2]);
if (resolved) out.push(resolved);
}
return out;
}
// Breadth-first from `entry`, returning the shortest chain of files that ends
// at one of the forbidden modules, or null when none is reachable.
function chainToForbidden(entry, forbidden) {
const seen = new Set([entry]);
const queue = [[entry]];
while (queue.length > 0) {
const chain = queue.shift();
for (const next of requiresOf(chain[chain.length - 1])) {
if (forbidden.has(next)) return chain.concat([next]);
if (seen.has(next)) continue;
seen.add(next);
queue.push(chain.concat([next]));
}
}
return null;
}
const rule = {
meta: {
type: "problem",
docs: {
description:
"the background bundle must not be able to reach the" +
" module-level state singleton in src/shared/state.js",
},
schema: [],
messages: {
reachable:
"The background must not reach the shared state singleton:" +
" {{chain}}. The MV3 worker never populates it, so reading it" +
" serves DEFAULT_STATE. Use getState()/updateState() from" +
" src/background/state.js instead.",
},
},
create(context) {
return {
"Program:exit"(node) {
const filename = context.filename;
// ESLint lints from the repo root, which is also where the
// forbidden paths are anchored.
const forbidden = new Set(
FORBIDDEN.map((module) =>
path.resolve(context.cwd, module),
),
);
const chain = chainToForbidden(
path.resolve(filename),
forbidden,
);
if (!chain) return;
context.report({
node,
messageId: "reachable",
data: {
chain: chain
.map((file) => path.relative(context.cwd, file))
.join(" -> "),
},
});
},
};
},
};
module.exports = {
rules: { "no-state-singleton-in-background": rule },
};

View File

@@ -0,0 +1,123 @@
// The modules a given entry point's bundle may not contain, keyed by the
// repo-relative entry point.
//
// ONE table, read by both layers that act on it: build.js asserts it against
// esbuild's own metafile (the guarantee), and
// script/lib/eslint/noStateSingletonInBackground.js reports the same
// prohibition in the editor (fast feedback). It lives here because a second
// literal copy of the path is exactly how a rename disarms one layer while the
// other still looks enforced.
//
// src/shared/state.js holds a module-level `state` object, loaded once by
// loadState() and mutated in place from then on. That is the popup's model:
// one page, one load at boot, one lifetime. The MV3 service worker has no
// "once" — it is killed when idle and revived by the next message, nothing
// loads state at module scope, and an unpopulated read was answered out of
// DEFAULT_STATE in silence. Five defects came from that, one of which
// destroyed a wallet (https://git.eeqj.de/sneak/AutistMask/issues/324). The
// background has its own per-call storage layer in src/background/state.js
// instead.
//
// What build.js's assertion covers, measured rather than assumed:
//
// - Any import of a listed module, at any hop, in any specifier syntax,
// however esbuild resolved it. The check reads the input list esbuild
// reported for the emitted bundle, so it is the resolution the shipped
// file was built from and not a model of it. Measured on a computed
// specifier that esbuild constant-folds (`require("../shared/" +
// "state")`), on a computed specifier it resolves as a glob
// (`import("../shared/" + variable)`), and on a symlink to the module
// (esbuild reports the real path): each is `make build` exit 2.
//
// - Every background entry point, whether or not anyone remembered to list
// it. A bundled entry point under BACKGROUND_ENTRY_PREFIX with no line in
// this table fails the build (assertNoForbiddenInputs()), so adding a
// second worker entry point is protected by default rather than protected
// only if the person adding it knew about this file. Measured: bundling
// src/background/worker2.js with no line here is `make build` exit 2.
//
// - NOT covered: a COPY of a listed module at another path. The table is
// keyed by path, so `cp src/shared/state.js src/shared/stateCopy.js` plus
// a background require of the copy is `make build` exit 0 and `make lint`
// exit 0 (measured). The copy carries the singleton's own guard, so
// defects 1-3 of https://git.eeqj.de/sneak/AutistMask/issues/324 — a read
// of a field nothing loaded — become a loud StateNotLoadedError instead of
// a silent DEFAULT_STATE. Defects 4 and 5 do NOT: a copy also carries
// loadState(), and a stale read several awaits after a load, or a load
// detaching the objects an in-flight handler is mutating, are silent over
// a LOADED singleton whether it is the original or a copy. So the residual
// is wider than "it fails loudly". A newly WRITTEN singleton has no
// backstop at all.
//
// - NOT covered: a background-behaving entry point outside
// BACKGROUND_ENTRY_PREFIX. The default protection above is keyed on that
// directory, which is also what eslint.config.js scopes the rule to, so a
// worker entry point placed somewhere else is covered by neither layer and
// needs its own line here.
//
// The ESLint rule's bounds are its own and are narrower: it matches specifiers
// textually, so a computed specifier and a symlink to a listed module are
// reported by the build and not by the rule. Both are pinned as non-reports in
// tests/backgroundStateLintRule.test.js and are `make build` exit 2 (measured).
// A second background entry point reached by one of those two shapes is
// therefore caught by the build and not by the rule — which is the same
// division of labour as everywhere else here, not an extra hole.
//
// Every way the table itself can rot is a failure rather than a quiet pass:
//
// - a KEY no bundled entry point matched, and a listed MODULE this build
// bundled nowhere: assertForbiddenTableCovered(), at the end of a build;
// - an entry that lists NO modules, and a table with no entries at all:
// assertTableWellFormed() below, at require time — so it fails the build
// and the lint run alike, because the rule reads the same values and an
// empty list leaves it with nothing to look for.
//
// All of it is pinned by tests/buildForbiddenInputs.test.js.
// What counts as a background entry point, and therefore must be listed above.
// The build has no other notion of one: entry points are the paths handed to
// bundle(), and this prefix is the narrowest rule that names the worker's
// directory. eslint.config.js scopes the lint rule with the same prefix, from
// this constant, so the two layers cannot disagree about what "background"
// means.
const BACKGROUND_ENTRY_PREFIX = "src/background/";
const FORBIDDEN_INPUTS = {
"src/background/index.js": ["src/shared/state.js"],
};
// Refuse a table that cannot prohibit anything. An entry whose module list is
// empty passes every check in both layers while enforcing nothing: the build
// finds no module to look for and records the entry as checked, and the rule's
// forbidden set — Object.values(...).flat() — comes back empty, so a plain
// `require("../shared/state")` in the worker is green everywhere. That is a
// one-character edit, so it fails here, where the table is defined and both
// layers must load it, rather than in either layer's own checks.
function assertTableWellFormed(table) {
const entries = Object.entries(table);
if (entries.length === 0) {
throw new Error(
"FORBIDDEN_INPUTS is empty, so nothing is prohibited anywhere. " +
"Removing the last entry disables the guarantee behind " +
"https://git.eeqj.de/sneak/AutistMask/issues/324.",
);
}
for (const [entry, modules] of entries) {
if (!Array.isArray(modules) || modules.length === 0) {
throw new Error(
`FORBIDDEN_INPUTS["${entry}"] lists no modules, so it ` +
`prohibits nothing while still looking enforced. Give it ` +
`the modules that entry point may not reach, or remove ` +
`the entry.`,
);
}
}
}
assertTableWellFormed(FORBIDDEN_INPUTS);
module.exports = {
BACKGROUND_ENTRY_PREFIX,
FORBIDDEN_INPUTS,
assertTableWellFormed,
};