When both preflights succeed, GetMigrationSummary now compares the URL
fields in the parsed SoundTouchSdkPrivateCfg.xml (read via SSH) against
the matching keys in `getpdo CurrentSystemConfiguration` (read via
telnet) and appends a Warnings entry for any field whose values differ.
The two sources can briefly disagree because `sys configuration …`
writes the runtime layer while envswitch writes the parallel persistence
layer and the on-device XML file is only re-rendered after a reboot.
The warning text says exactly that, so the UI can surface a non-fatal
hint instead of treating a freshly-migrated-but-not-yet-rebooted device
as broken.
Adds Warnings []string on MigrationSummary, parseGetpdoConfig (a
key=value parser tolerant to banner/prompt noise), and
crossCheckPreflights wired in as step 9 of GetMigrationSummary.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds Manager.isTelnetMigrated, which substring-matches m.ServerURL's
hostname against TelnetVerifiedConfig — the response captured by the
preflight's `getpdo CurrentSystemConfiguration`. Mirrors the existing
isXMLMigrated semantics so users see consistent migration-state
detection regardless of which transport the device exposes.
checkIsMigrated no longer early-returns on !SSHSuccess. Telnet runs
first and unconditionally; the SSH-based hosts/resolv.conf checks still
run when SSH is reachable, since neither variant shows up in
`getpdo CurrentSystemConfiguration`. This closes the gap where a
USB-unlock-refusing speaker (SA-5, ST520, recent ST Portable) that had
already been migrated via telnet was silently reported as IsMigrated:
false in the UI.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
GetMigrationSummary now kicks off telnetPreflight in a goroutine at
entry and merges the four Telnet* fields into the main summary just
before returning. Wall time becomes max(ssh, telnet); the two transports
are queried independently and their results combined — SSH retains
visibility into /etc/hosts, /etc/resolv.conf and the on-device XML
config, while telnet contributes the live URL set readable via
`getpdo CurrentSystemConfiguration` without root.
Race-free by construction: the goroutine writes to its own
MigrationSummary instance and only the four telnet fields are copied
back. Verified with `go test -race`.
Tests cover telnet-only, ssh-only, and both-succeed paths.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds Manager.telnetPreflight that dials port 17000, captures the banner,
and runs `getpdo CurrentSystemConfiguration` to read back the device's
live URL configuration. Errors are recorded on TelnetProbeError instead
of returned, so the probe is best-effort and never breaks summary
construction.
This is the data-gathering layer that the four already-declared
TelnetReachable / TelnetBanner / TelnetVerifiedConfig / TelnetProbeError
fields on MigrationSummary were waiting for. Subsequent iterations wire
the preflight into GetMigrationSummary (in parallel with SSH) and use
TelnetVerifiedConfig as a SSH-free signal for "already migrated".
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The previous filepath.IsLocal-up-front pattern in safeJoin/safeJoin-equivalents
turned out not to satisfy CodeQL's go/path-injection rule — the post-validation
filepath.Join still constructs the joined string from tainted input, so the
analyser conservatively assumes the os.* sink that consumes it is tainted
too. Only one of 34 alerts closed on the previous attempt.
Switch to *os.Root (Go 1.24+, available on the project's 1.26.3 toolchain).
The Go runtime guarantees that operations on a Root cannot escape the
anchored directory regardless of what's in the relative path, and CodeQL has
a built-in model that recognises *os.Root.* methods as path-traversal
sanitisers. Result: every os.* sink in the datastore, marge, recorder,
mirror parity-mismatch writer, and docs handler is now reached only via a
*os.Root, which closes the rule-level alerts cleanly.
Changes per file:
* pkg/service/datastore/datastore.go — Adds a `root *os.Root` to DataStore,
lazily opened at first use (after MkdirAll-ing baseDir) and closed by a
new `(*DataStore).Close()`. Adds package-private helpers
(rootStat / rootReadFile / rootWriteFile / rootMkdirAll / rootRemove /
rootRemoveAll / rootRename / rootReadDir / rootOpen / rootExists) plus
three exported wrappers (ReadDirUnderBase, MkdirAllUnderBase,
WriteFileUnderBase) for the cross-package marge / handlers callers.
Every os.* call that previously consumed safeJoin output now goes through
these helpers. The post-join belt-and-suspenders prefix check inside
safeJoin is preserved as a defence-in-depth fallback.
* pkg/service/marge/marge.go — Replaces the five `os.ReadDir(devicesDir)`
call sites with `ds.ReadDirUnderBase(...)` so the datastore's root
enforces containment.
* pkg/service/proxy/recorder.go — Mirrors the datastore pattern with its
own `root *os.Root` anchored at Recorder.BaseDir, lazily opened. New
helpers convert the eight existing `os.*` sites that consume sessionID
/ relPath / sanitizedSegments inputs. The earlier safeJoin (filepath.IsLocal
pre-check) stays in place as the same belt-and-suspenders guard.
* pkg/service/handlers/handlers_docs.go — Opens a *os.Root at "docs" via
sync.Once and reads file content (and SUMMARY.md sidebar) through it.
Removes the prior filepath.IsLocal pre-check; the runtime now guarantees
containment.
* pkg/service/handlers/mirror_middleware.go — Routes the parity-mismatch
JSON write through `s.ds.WriteFileUnderBase` so the datastore's root
performs the path-traversal sanitiser.
Behavioural fix: *os.File.ReadDir(-1) returns directory entries in
filesystem order, but os.ReadDir is documented to sort by name and at least
one regression test
(handlers.TestMargeAccountFullExcludesEmptyAmazonSource) depends on the
sorted contract. Both rootReadDir helpers explicitly sort by name to match.
All test suites pass for the touched packages; the unrelated
TestDocsConsistency failure about untracked working-tree docs is
pre-existing. golangci-lint reports 0 issues.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Wire up X-Real-IP / X-Forwarded-For / True-Client-IP support for
deployments fronted by a reverse proxy, while staying safe on flat-LAN
deployments where a malicious speaker could spoof those headers
directly.
Two new fields on `datastore.Settings`:
* TrustForwardedHeaders (bool, default false) — opt-in switch.
* TrustedProxyCIDRs ([]string, default `["127.0.0.0/8", "::1/128"]`)
— only requests whose immediate TCP peer falls in one of these
blocks may have their source IP rewritten from forwarded headers.
Loopback default matches the documented same-host nginx layout in
docs/guides/HTTPS-SETUP.md.
New middleware in `pkg/service/handlers/middleware_realip.go`:
* TrustedRealIP wraps `chi/middleware.RealIP` with a trusted-peer
gate. When the immediate TCP peer is in the allowlist, chi's
parsing handles the actual header → IP rewrite. When it isn't
(e.g. a speaker sending forwarded headers itself), we ignore the
headers and r.RemoteAddr stays as-is.
* ParseTrustedProxyCIDRs converts string CIDRs into *net.IPNet,
applying the loopback default on empty input and erroring loudly
on invalid entries.
Server.TrustedRealIPMiddleware() returns the middleware (or nil) by
reading the live settings; the router setup in
cmd/soundtouch-service/main.go installs it as the very first
middleware so SnapshotMiddleware and downstream handlers see the
correct r.RemoteAddr.
HandleMargePowerOn now prefers r.RemoteAddr over the body's
self-reported `<IPAddress>` for outbound credential push:
* The body field is treated as a hint only — a malicious LAN speaker
could set it to any value; using it for outbound HTTP requests is
the SSRF surface the previous zeroconf hardening was guarding
against from the sink side. Fixing it at the source as well closes
the gap entirely.
* When body IP and TCP source disagree, a log line names both and
the device ID so the discrepancy is investigable.
* RemoteAddr is unparseable → fall back to the body so we don't
silently drop the priming.
docs/guides/HTTPS-SETUP.md gains a follow-up note next to the existing
nginx snippet explaining the new flag, the loopback-only default, and
the explicit warning against enabling the flag on a flat-LAN
deployment without a real proxy.
Eleven test cases in middleware_realip_test.go lock in the gate
behaviour: trusted peers honoured for X-Real-IP / X-Forwarded-For /
no-headers / IPv6, untrusted peers' headers ignored, garbage values
rejected, ParseTrustedProxyCIDRs covers default / override / invalid.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Building on the strict literal-IP validator from the previous commit,
make the runtime error self-explanatory so anyone tripping on a
hostname URL can fix it in one shot:
* Errors now lead with the offending zeroconf URL and the rejected
host, so wrapping by GetInfo / PushCredentials / pushSimplifiedToken
doesn't bury the actual bad value.
* The "host must be a literal IP" error suggests two concrete one-liner
resolutions (`getent hosts <name>` and `dig +short <name>`) so the
user has a copy-paste fix.
* The "host is not on a local network" error names the accepted ranges
(loopback / RFC1918 private / link-local v4+v6) so the user knows
what they're allowed to pass.
docs/guides/SOUNDTOUCH-SERVICE.md gains a bullet under Security
Considerations explaining the constraint and the rationale (LAN-resident
SSRF surface), so the strict behaviour is documented rather than a
surprise.
The 17 TestValidateZcBaseURL cases still pass — only the message bodies
changed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CodeQL re-fired three new go/request-forgery alerts (#134/135/136) on
the lines my previous validateZcBaseURL refactor introduced. The
previous validator accepted hostname-style hosts unchanged, so even
though the IP-class check ran when applicable, u.String() at the call
sites still emitted the original tainted host into the request URL —
which is exactly what CodeQL traces.
Tighten validateZcBaseURL to:
* require the host to parse as a literal IP — DNS / mDNS hostnames
are rejected (with a clear error explaining the caller should
resolve to a private IP first); doing the lookup inside the
validator would re-introduce the SSRF surface CodeQL is flagging,
because malicious DNS could point a *.local name at a public host
between the lookup and the request.
* require that IP to be loopback / RFC1918 private / IPv4-or-IPv6
link-local. Anything else (global IPs in either family) is refused.
* rebuild the returned *url.URL from validated components — scheme
(already checked), the validated IP literal joined with the
original port, and the original path. Pre-existing query/fragment
are stripped so callers attach their own ?action= cleanly. CodeQL
recognises this fresh-construction pattern as taint sanitisation.
In practice this matches what SoundTouch speakers actually announce:
IP-based zeroconf URLs at port 8200 against an LAN address. The
existing PushCredentials_FullRoundTrip and FallbackOnGetInfoFailure
tests already exercise the loopback path through httptest.NewServer
and pass unchanged.
Adds TestValidateZcBaseURL covering 17 inputs — 9 accept (loopback,
private 10/172/192, link-local v4, IPv6 loopback, IPv6 link-local,
strips query) and 8 reject (public IPv4, public IPv6, hostname,
plain hostname, ftp/file schemes, empty host, unparseable) — to lock
the new contract in.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The previous commit made credential-header redaction unconditional in
proxy log output, which is the right safety floor for production but
inconvenient for local debugging when a developer wants to inspect
Authorization / Cookie / X-Bose-Token values flowing through the
service.
Add an explicit "I-know-what-I-am-doing" toggle:
* New LoggingProxy.UnsafeLogCredentialHeaders bool field.
* Default off — the redaction floor stays in place.
* Reads the LOG_PROXY_CREDENTIALS env var so a developer can flip it
on without recompiling, mirroring the existing LOG_PROXY_BODY
pattern.
* When true, formatHeaders skips both the always-sensitive floor and
the broader Redact policy, so log lines contain raw header values.
CodeQL's go/clear-text-logging rule continues to be satisfied because
the default code path still redacts; only an explicit opt-in via
configuration produces unredacted output, mirroring how
AllowInsecureUpstreamTLS works.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CodeQL alerts #121, #122, #123 (go/request-forgery) flagged the three
client.Get / client.PostForm sites in pkg/service/zeroconf/zeroconf.go
that build their request URL by string-concatenating the caller-supplied
zcBaseURL with "?action=…". The base URL ultimately originates from a
device-pairing payload that the speaker pushes to us, so unvalidated
input could redirect outbound HTTP requests to arbitrary hosts (server-
side request forgery).
Add validateZcBaseURL which:
* parses zcBaseURL via net/url so the scheme and host are first-class
values rather than substrings,
* requires the scheme to be http or https,
* rejects literal IP hosts that aren't loopback / RFC1918 private /
link-local — those are the only places a real SoundTouch speaker
can live on a local network, and a global IP would be an obvious
exfiltration target,
* leaves hostname-style hosts (e.g. mDNS *.local) accepted: name
resolution itself is a separate trust boundary on the local segment.
A small withAction helper builds the per-call URL from the validated
base URL via url.Values rather than string concatenation, which CodeQL
recognises as a non-tainted construction.
GetInfo, PushCredentials and pushSimplifiedToken each call
validateZcBaseURL up-front so all three CodeQL alerts close in a
single pass. PushCredentials also re-validates even though it then
calls GetInfo (which validates again) so the fallback to
pushSimplifiedToken on getInfo failure is also gated.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CodeQL alerts #70 and #71 (go/disabled-certificate-check) flagged the
hard-coded `InsecureSkipVerify: true` in handlers_proxy.go (the
/proxy/{url} reverse proxy) and mirror_middleware.go (the parity-check
mirror). Both target *.bose.com whose certificate chain is becoming
unreliable post end-of-service, but unconditionally disabling
verification is still wrong: a deployment that doesn't actually need
the bypass loses TLS hygiene for free.
Add an `AllowInsecureUpstreamTLS bool` field to datastore.Settings,
default false. Read it in both call sites — they aren't on a hot path
— and pass the value as InsecureSkipVerify. CodeQL accepts the
configurable boolean as a non-flag (vs. the previously hard-coded
`true`), and the runtime behaviour now defaults to verifying
certificates with an explicit opt-in for the broken-chain scenario.
Behaviour change: TLS upstream traffic is verified by default. Anyone
relying on the previous always-skip behaviour can re-enable it by
setting `"allow_insecure_upstream_tls": true` in settings.json.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CodeQL alert #43 (go/clear-text-logging) flagged that headers flow to
log.Printf in pkg/service/proxy/proxy.go. The existing implementation
only redacted when LoggingProxy.Redact was true — an opt-in. CodeQL is
right to flag this: the safety floor for credential-bearing headers
should not depend on caller configuration.
Split the sensitive-header list into two:
* alwaysSensitiveHeaders — Authorization, Proxy-Authorization, Cookie,
Set-Cookie, X-Api-Key, X-Bose-Token. Redacted unconditionally,
regardless of LoggingProxy.Redact.
* sensitiveHeaders — kept as a compatibility alias pointing at the same
list, and still gated on Redact for any future use cases that want
*additional* opt-in redaction beyond the floor.
Behaviour change is strict tightening: nothing that was previously
hidden becomes visible, and credentials that would have been logged
when Redact was false are now hidden by default.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CodeQL flagged five reflected-XSS sites where caller-supplied query
parameters or path segments were concatenated into HTML responses
without escaping:
* handlers_mgmt.go:147 — Spotify oauth error landing page
* handlers_mgmt.go:490 — Amazon oauth error landing page
* handlers_docs.go:65 — <title> built from r.URL.Path
* recorder_middleware.go:83, mirror_middleware.go:205 — passthrough
Write()s carrying tainted bytes from the three sources above
Wrap each user-controlled value in html.EscapeString before it lands
in the HTML body. The escaped output covers the upstream sources so
the middleware passthrough alerts close as well.
For handlers_docs the rendered markdown (`output`) and sidebar are
server-controlled (loaded from on-disk doc files) and intentionally
contain HTML, so only the URL path is escaped — the documentation
content itself still renders normally.
Handler test suite passes; pre-existing TestDocsConsistency failure
about untracked working-tree docs is unrelated.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CodeQL flagged 34 go/path-injection alerts across datastore.go, marge.go,
recorder.go, handlers_docs.go and mirror_middleware.go. The existing
defences (DataStore.safeJoin's post-join prefix check, handlers_docs's
HasPrefix(filepath.Clean(...))) are functionally correct but sit
downstream of the join, so CodeQL's interprocedural taint tracking
treats every os.* sink that consumes them as still tainted.
Move the validation up-front using filepath.IsLocal, which CodeQL
recognises as a path-traversal sanitiser. IsLocal rejects absolute
paths, ".." segments, and (on Windows) reserved device names — the
same set the existing checks intended to block, just expressed in the
shape the analyser understands.
Changes:
* DataStore.safeJoin (datastore.go) — pre-validates each non-empty
element with filepath.IsLocal before joining. Existing post-join
prefix check stays as belt-and-suspenders. ~30 of the 34 alerts
flow through this helper.
* Recorder (recorder.go) — adds a new (*Recorder).safeJoin method
with the same sanitiser. getRecordingDir, DeleteSession,
GetInteractionContent and ArchiveSession route through it; their
signatures already returned error so plumbing it through is local.
* HandleDocs (handlers_docs.go) — replaces the post-join HasPrefix
check with an up-front filepath.IsLocal gate.
* Mirror parity recorder (mirror_middleware.go) — also strips
backslash separators (Windows) and gates the resulting filename
component on filepath.IsLocal, falling back to "invalid" rather
than letting malformed paths reach os.WriteFile.
No behaviour change for legitimate inputs (account IDs, device IDs,
session IDs, doc paths all satisfy IsLocal). Datastore and proxy
test suites pass; handler suite's pre-existing TestDocsConsistency
failure is unrelated.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Sweeps the remaining instances of the same pattern that triggered CodeQL
alert 132 in PR #240's review: status messages built by string-concatenating
the user-controlled `display` (device name) into `.innerHTML`. None of
these had ever needed HTML formatting; they're all plain status text.
Converts 26 sites across reboot(), revert(), migrate(), showSummary(),
trustCA(), ensureRemoteServices(), removeRemoteServices(), backup(),
plus fetchDevices' error fallback and the loadAccount sync log line.
The one site that genuinely needs intentional <strong> formatting — the
migrate() success message ("Please reboot the device to activate the
changes.") — is rebuilt with replaceChildren + createElement so the
device name still flows through createTextNode rather than HTML parsing.
Out of scope (intentionally left for a separate pass): the dashboard
table rows, account-metadata templates, and the error.message-into-
colored-span / redirectUrl-into-href patterns. Those are different
classes and benefit from a focused refactor.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CodeQL alert 132 flagged the reboot status line as a sink that received
user-controlled DOM text (device names from the migration/sync select
options and table rows) without escaping. Six data-flow paths converged
on script.js:1950.
Switch the sink at line 1950 from .innerHTML to .textContent — the
status message has never needed HTML formatting. The pre-existing
display-into-innerHTML pattern still exists elsewhere in this file but
those lines aren't in this PR's scope and are tracked by their own
historical alerts.
Also harden the (newer) `currentP.innerHTML = ... <strong> + data.current
+ </strong> ...` line in loadAccountIDSuggestions: rebuild the paragraph
with replaceChildren + createElement so the account ID never becomes
HTML, even though it's expected to be a 7-digit string.
Coerce known account IDs to String() when populating the existing-account
dropdown so the IDE's type inference stops complaining about
opt.value = id; / opt.textContent = id; on data of unknown[] type.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The previous copy said "see the panel below" while the Pair Account panel
is intentionally hidden until migration succeeds (loadAccountIDSuggestions
makes it visible). Reword so users know the panel will appear after they
click Migrate.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Migration dropdown gains a "Telnet (Port 17000) — no SSH required" option
and drops the deprecated /etc/hosts entry from the visible choices. The
hosts code path still exists in the backend for now; it is just no longer
reachable through the UI.
* New `telnet-method-pane` shows a brief explanation, the HTTP-only
limitation, and a hint that pairing may be required after migration.
* New `pair-account-pane` (initially hidden) renders three controls:
- dropdown of accounts already in the local datastore (so a fresh device
can be re-attached to an existing account),
- 7-digit input field with HTML pattern validation,
- a Generate button that picks a random non-colliding 7-digit ID.
When :8090/info already exposes a margeAccountUUID the panel pre-fills
it and offers to keep it; otherwise the device is treated as fresh.
* `pairAccount(deviceId)` POSTs to /setup/pair-account/{deviceId} with the
selected ID and surfaces the breadcrumb (HTTP vs telnet fallback) in the
status line.
* `reboot()` now passes ?method=telnet|ssh, derived from the migration
method dropdown (telnet for telnet, ssh otherwise) so a device that was
migrated without SSH access can also be rebooted without SSH access.
* After a successful telnet migration, `loadAccountIDSuggestions` runs
automatically so the user is led straight into the pairing step.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds an SSH-free third migration path that drives the SoundTouch device's
diagnostic shell on TCP port 17000, plus a hardened replacement for the
fragile /setMargeAccount HTTP pairing call.
* `pkg/telnet` — new reusable, dependency-free client (sibling of `pkg/ssh`)
with deadline-driven Dial / Probe / SendCommand / Close. Mock-server tests
cover happy path, command-not-found, mid-stream close, and the wedged-device
read-timeout scenario.
* `setup.MigrationMethodTelnet` — runs `sys configuration` for all four URLs
plus the parallel `envswitch boseurls set` persistence layer that otherwise
wins on reboot, then verifies with `getpdo CurrentSystemConfiguration`.
Aborts on the first non-OK response so configuration is never half-written.
No SSH backup or rw pre-flight (the path is SSH-free by design).
* `setup.PairAccount` — probes :8090/supportedURLs first, time-bounds
POST /setMargeAccount aggressively (5s connect / 12s total) to avoid the
hangs reported in #236, and falls back to `envswitch accountid set <id>`
over telnet when the HTTP endpoint is missing or wedged. Returns a
PairAccountResult breadcrumb so the UI can show which path actually
succeeded.
* `setup.Reboot(deviceIP, method)` — gains a RebootMethod selector;
RebootMethodSSH stays the default (preserving prior behavior),
RebootMethodTelnet sends `sys reboot` over a fresh telnet session and
treats the inevitable socket-close as success.
* New endpoints on `/setup`:
- GET /account-id-suggestions/{deviceId} — returns the device's current
margeAccountUUID (from :8090/info) plus known account IDs from the
datastore, so the UI can offer reuse.
- POST /pair-account/{deviceId}?account_id=NNNNNNN — invokes PairAccount;
the existing reboot endpoint reads ?method=ssh|telnet from the query
string.
* Helpers `IsValidAccountID` (exactly 7 digits) and `GenerateAccountID`
(crypto/rand, retries on collision against a known-IDs list).
Documentation in docs/analysis/TELNET-MIGRATION-METHOD.md is updated to match
the implementation: bare-URL convention for `soundtouch-service`, no automatic
`sys reboot` (user-initiated via the existing button with a method selector),
and the realised package layout. The /etc/hosts method is intentionally not
exposed in the new flow.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The on-device Sources.xml carries only displayName + sourceKey for AUX,
no id and no type. The previous read path synthesized id="2000001+i" and
type="AUX" (echoed from SourceKey.Type), which the speaker rejects as
INVALID_SOURCE once it pulls config from soundtouch-service after
migration. Look up known providers in getDefaultSources and fill
canonical id/type/sourceproviderid; also drop the AUX carve-out in
marge's ensureSourceType so existing poisoned type="AUX" entries are
normalized to type="Audio" at the served-XML layer.
Relates to https://github.com/gesellix/Bose-SoundTouch/issues/195
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Why: Operators need to control device discovery from the command line
without touching the persisted settings file, and a zero discovery
interval should be unambiguously off rather than running an
immediate-fire scan loop.
- Add --discovery-enabled BoolFlag (default true, env DISCOVERY_ENABLED)
and thread it through serviceConfig, applyPersistedSettings, and
createDefaultSettings so CLI/env can seed initial state and persisted
settings still take precedence on subsequent runs.
- HandleUpdateSettings now forces discoveryEnabled=false whenever the
resulting discoveryInterval is zero.
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The speaker rejects /select with source="AUX" and an empty sourceAccount
as INVALID_SOURCE, so the audio path never reaches APAuxSrc. Default the
sourceAccount to "AUX" inside SelectSource and align ItemName to "AUX
IN" to match the device's own button-press payload.
Relates to https://github.com/gesellix/Bose-SoundTouch/issues/195
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The OverrideSdkPrivateCfg.xml override path introduced in #209 does not
work on SoundTouch 10 (and likely other models): the firmware ignores
the override file, leaving the device pointing at the original Bose
cloud URLs. Revert to editing SoundTouchSdkPrivateCfg.xml directly with
a .original backup, which is the approach known to work.
Relates to #214
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
client.Run uses CombinedOutput, so when
`/mnt/nv/OverrideSdkPrivateCfg.xml` is absent (the default for devices
migrated with pre-0.71.0 code) the cat stderr is returned as the
override config and surfaced to the migration page UI as "Current Config
(on Speaker)". Gate the branch on `[ -f ... ]` first, mirroring the
legacy .original check.
Relates to #209
Relates to #214
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Store the user's email address (not Amazon account ID) in
sourceKey.account and set source type to "AMAZON" so the speaker
firmware recognises Amazon Music sources the same way as the original
Bose cloud.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds a "Download CA Certificate" button in the Settings tab
(system-level convenience for importing the cert into browsers, curl,
Python clients, etc.) and a "Download CA cert" link next to the existing
"Trust CA Now" button in the Migration tab. Both link to the existing
/setup/ca.crt endpoint.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Use /mnt/nv/OverrideSdkPrivateCfg.xml (the firmware's override path)
rather than editing /opt/Bose/etc/SoundTouchSdkPrivateCfg.xml directly.
A malformed override cannot cause a reboot loop because the device falls
back to the untouched original.
Revert now removes the override file; legacy .original backups are still
restored for devices migrated with older code. checkCurrentConfig reads
the override path first so IsMigrated detection works correctly with the
new approach.
Credit: Ueberbose team, discovered via [soundcork
documentation](https://github.com/deborahgu/soundcork#configuring-the-bose-speaker-to-use-the-soundcork-server).
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
domains[0] was non-deterministic (Go map iteration) and could resolve to
any domain in the list including Bose-owned domains. Adds CommonName field
to CertificateManager, defaulting to "localhost", set to the device hostname
at startup. All Bose domains remain in the SAN where clients actually look.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
RSA-4096 CA generation blocks service startup for minutes on slow ARM
hardware. The CA key is only used to sign server certs, never in TLS
handshakes, so 2048 bits provides sufficient security for a local CA
while being ~4-8x faster to generate.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- resolveIP now returns (string, error): error when result did not come
from the device's own SSH ping (service-side fallback or total
failure)
- migrateViaResolvConf and parseTargetURLAndResolveIP abort on error,
preventing a bad IP from being written to the device
- GetMigrationSummary captures the error in ResolveIPError and falls
back
to the hostname for the preview display; XML migration is unaffected
- Web UI shows a warning box with the error and a docs link when
resolution
is uncertain; migrate button stays enabled for the XML method
- Add hostname resolution troubleshooting section to TROUBLESHOOTING.md
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
- Rename proxy option values: 'upstream' → 'proxied', 'official' →
'original'
- Add 'original' option to preserve current device URL as-is per field
- Drop proxyURL guard in applyProxyOptions so 'original' works without a
proxy
- Abort migration if on-device backup cannot be created (was
warning-only)
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Registers HandleAlexaCertificate on POST /alexa/certificate. The handler
logs the device MAC from the request body and returns 501 Not
Implemented with a JSON error explaining that AWS IoT integration is
required to provision Alexa device certificates.
Adds voice.api.bose.io to both /etc/hosts domain lists in setup.go (DNS
intercept was already covered by the bose.io wildcard entry in dns.go).
Relates to https://github.com/gesellix/Bose-SoundTouch/discussions/84
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Endpoints:
- POST /streaming/music/musicprovider/{id}/trial/is_eligible (reuses
is_eligible handler)
- POST /bmx/tunein/v1/favorite/{stationID} with datastore persistence
(SaveTuneInFavorite)
- DELETE /bmx/tunein/v1/favorite/{stationID} (DeleteTuneInFavorite)
- POST /bmx/core02/svc-bmx-adapter-orion/prod/orion/token (anonymous
Orion token)
- GET /bmx-icons/* serving embedded static/media assets (media.bose.io)
- GET /ced/* serving embedded firmware index, release notes, and 10
app-help XMLs (downloads.bose.com)
Add media.bose.io and downloads.bose.com to DNS redirect lists (setup.go
both domain slices, dns.go shouldIntercept list, main.go getDomains
map). Document implemented endpoints in
tests/interactions_20260502_missing_external.md; mark rows 0246–0247 as
self/☑ in the interactions table.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Introduces a standalone `soundtouch-backup` CLI with three subcommands:
- `all`: authenticates with the Bose cloud, backs up account data, then
reads device IPs from devices.xml and backs up each reachable speaker
- `cloud`: fetches account profile, devices, sources, presets, and full
endpoint from streaming.bose.com
- `local`: backs up each speaker via HTTP API (12 endpoints) and
optionally via SSH (individual files + /opt/Bose/etc/ and
/mnt/nv/BoseApp-Persistence/1/ directories)
Also centralises pkg/service/ssh → pkg/ssh so both the service and the
backup tool share the same SSH client; adds ReadFile and ReadDir
methods, and handles the firmware quirk where cat exits 1 on empty
files.
Output is a single dated .tar.gz or .zip archive.
Example flow:
```shell
gesellix@Mac Bose-SoundTouch % go run ./cmd/soundtouch-backup all --output _/cloud-backup --email user@example.com
Password:
Authenticating as user@example.com...
✓ Authenticated (account ID: 1234567)
✓ email address (107 bytes)
✓ devices (1492 bytes)
✓ sources (1111 bytes)
✓ presets (2585 bytes)
✓ full account (55037 bytes)
Found 2 device(s) in cloud account, attempting local backup...
✓ ST20: 12 files via HTTP
⚠ ST20: SSH skipped /etc/remote_services (Process exited with status 1)
⚠ ST20: SSH empty file /mnt/nv/remote_services
✓ ST20: 64 files via SSH
✓ ST10: 12 files via HTTP
⚠ ST10: SSH empty file /etc/remote_services
⚠ ST10: SSH skipped /mnt/nv/remote_services (Process exited with status 1)
✓ ST10: 48 files via SSH
Archive written: _/cloud-backup/soundtouch-backup-2026-05-02.tar.gz (141 files)
```
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Attempting to request amazon_music:access with a standard application
client ID (amzn1.application-oa2-client.*) returns HTTP 400
lwa-invalid-parameter-bad-scope from the LWA authorization endpoint.
The scope is gated to Amazon Music partner device client IDs.
Revert scope to "profile" (working state) and document the confirmed
blocker with the exact error. Path forward: Amazon Music partner
registration for a device client ID; one-line change to AmazonScopes
when available.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Amazon bridge: fall back to sync/legacy on any error from
SetMusicServiceOAuthAccount (not only error 1029); timeouts from
unresponsive speakers no longer silently skip the fallback chain
- Amazon bridge: reduce speaker client timeout from 30s to 5s for
faster failure on local network calls
- marge: resolveSourceName now prefers SourceName/DisplayName over
SourceKeyAccount, so Amazon (and Spotify) sources show the account
holder's name instead of the raw account ID
- docs: update amazon-music-oauth.md with real-world test results;
music-api.amazon.com returns 401 because standard LWA apps lack
music::* partner scopes — infrastructure is complete but streaming
is blocked pending Amazon partner access
- docs: add SELF-HOSTING.md and MUSIC-SERVICES.md user guides; link
both in SUMMARY.md
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add SpotifyClientID/Secret/RedirectURI and AmazonClientID/Secret/RedirectURI
fields to datastore.Settings for persistent storage
- Server: add amazonClientID/Secret/RedirectURI fields, SetAmazonConfig,
GetSpotifyConfig/GetAmazonConfig, ReinitSpotifyService/ReinitAmazonService,
and applyMusicServiceCredentials (called under lock from HandleUpdateSettings)
- GET /setup/settings: expose credential fields; mask secrets as "***" when set
- POST /setup/settings: apply credential updates and reinitialize services live
- applyPersistedSettings: fill in music credentials from settings.json when not
set via CLI/env (CLI takes precedence)
- Settings tab: replace read-only Spotify status with editable Client ID / Secret /
Redirect URI inputs for both Spotify and Amazon; save via existing Save button
- script.js: populate and collect the six new fields in fetchSettings/updateSettings
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Mirror the Spotify equivalents: pkg/testutils/amazon/handlers.go provides
HandleToken and HandleProfile for use in unit tests; tests/integration/mocks/amazon.go
wraps them in an AmazonMock with TokenURL() and ProfileURL() accessors.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add GetAccountByRefreshToken to amazon.Service — the speaker sends
the bare Atzr| refresh token (extracted from AmazonSecret JSON), not
a surrogate, so lookup must match against Account.RefreshToken
- Add amazonService field, SetAmazonService and IsAmazonConfigured to
Server (step 5 essentials required by the handler)
- Replace HandleBoseAmazonToken 501 stub with full implementation:
lookup by refresh token → RefreshAccessToken; fallback to
GetFreshToken; fallback to HandleBoseProxy if no service configured;
scope intentionally omitted from response
- Add handler tests covering the by-refresh-token path (mock LWA
server), the default-account path, and the no-service fallback
- Unlock assertions in post_oauth_token_amazon.http integration test
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Extract DH key exchange crypto from pkg/service/spotify into new
pkg/service/zeroconf package with exported functions and
AuthTypeOAuthToken constant (both Spotify and Amazon use auth type 4)
- Reduce pkg/service/spotify/zeroconf.go to thin wrappers around the
shared package; public API (PushSpotifyCredentials, ZeroConfGetInfo)
is preserved
- Add pkg/service/amazon package mirroring the Spotify service with
Amazon-specific differences: LWA endpoints, POST body credentials
(not Basic Auth), user_id/name profile fields, amazon/accounts.json
- Add PushAmazonCredentials delegating to shared zeroconf.PushCredentials
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Recognize Amazon Music in learned sources (classifyAsAmazon) and
AddSource dispatch, using CredentialTypeToken (cs1) not cs3
- Exclude Amazon from default sources: an empty-credential Amazon entry
triggers the speaker's AmazonController to fail JSON parsing with
MUSIC_SERVICE_ACCOUNT_LOGIN_FAILED; Amazon must only appear once a
real OAuth token is present
- Merge missing defaults into stored sources at request time so devices
with older Sources.xml still receive all current defaults
- Fix source providers ETag: was time.Now().UnixMilli() (always new),
now a content hash so If-None-Match/304 works correctly
- Include default sources fingerprint in GetETagForAccount so adding a
new default invalidates cached /full responses on speakers
- Refactor createLearnedSource into classifyLearnedSource +
classifyAsX helpers to reduce cyclomatic complexity below linter limit
- Add regression test for two-device scenario matching production setup
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Replace the simplified tokenType=accesstoken push with the full Spotify
Connect ZeroConf protocol: GET getInfo to fetch the speaker's 768-bit DH
public key, derive AES-128-CTR + HMAC-SHA1 keys from the shared secret,
and POST an encrypted LoginCredentials protobuf blob. Speakers that
receive a proper blob can self-refresh their Spotify session
independently, eliminating the need for periodic re-priming on token
expiry. Falls back to the raw token approach automatically when getInfo
fails, preserving compatibility with older firmware.
SHA1 is mandated by the Spotify Connect ZeroConf protocol spec for DH key derivation. This cannot be changed without breaking protocol compatibility.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Groups (stereo pairs of ST10 speakers) were read-only — the GET endpoint
always returned an empty <group/>. Add POST /account/{account}/group,
POST /account/{account}/group/{groupId}, and DELETE
/account/{account}/group/{groupId} with datastore persistence, matching
the API shape observed in soundcork. The GET endpoint now reads live
group state from the datastore.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
feat: log [UNHANDLED] for routes with no local handler
Every request that falls through to HandleNotFound now emits an
[UNHANDLED] METHOD path log line, making it immediately visible when a
speaker calls an endpoint we have not implemented. When proxyLogBody is
enabled the request body is also included (truncated to 512 bytes) and
restored before forwarding, so the proxy still sees the full payload.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Naive string concatenation (`rawURL + "&render=json"`) produced
malformed URLs when the input had no query string yet, or already
contained render=json. Replace with tuneInRenderJSONURI which parses and
sets the parameter cleanly. Also fix TuneIn search query encoding in the
self link and section href, and replace the http-prefix check for OPML
URIs with a proper host comparison.
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Tobias GesellchenandClaude Opus 4.7
