[feature] Discover webfinger through host-meta (#1588)

* [feature] Discover webfinger through host-meta

This implements a fallback for discovering the webfinger endpoint in
case the /.well-known/webfinger endpoint wasn't properly redirected.
Some instances do this because the recommendation used to be to use
host-meta for the webfinger redirect in the before times.

Closes #1558.

* [bug] Ensure we only ever update cache on success

* [chore] Move finger tests to their own place

This adds a test suite for transport and moves the finger cache tests
into there instead of abusing the search test suite.

* [chore] cleanup the test a bit more

We don't really need a separate function for the oddly located webfinger
response as we check the full URL string anyway

* Address review comments

* [chore] update config example

* [chore] access DB only through state in controller

internal/transport/controller.go

@@ -32,9 +32,9 @@ import (
 	"github.com/superseriousbusiness/activity/pub"
 	"github.com/superseriousbusiness/activity/streams"
 	"github.com/superseriousbusiness/gotosocial/internal/config"
-	"github.com/superseriousbusiness/gotosocial/internal/db"
 	"github.com/superseriousbusiness/gotosocial/internal/federation/federatingdb"
 	"github.com/superseriousbusiness/gotosocial/internal/log"
+	"github.com/superseriousbusiness/gotosocial/internal/state"
 )
 
 // Controller generates transports for use in making federation requests to other servers.
@@ -47,7 +47,7 @@ type Controller interface {
 }
 
 type controller struct {
-	db        db.DB
+	state     *state.State
 	fedDB     federatingdb.DB
 	clock     pub.Clock
 	client    pub.HttpClient
@@ -57,14 +57,14 @@ type controller struct {
 }
 
 // NewController returns an implementation of the Controller interface for creating new transports
-func NewController(db db.DB, federatingDB federatingdb.DB, clock pub.Clock, client pub.HttpClient) Controller {
+func NewController(state *state.State, federatingDB federatingdb.DB, clock pub.Clock, client pub.HttpClient) Controller {
 	applicationName := config.GetApplicationName()
 	host := config.GetHost()
 	proto := config.GetProtocol()
 	version := config.GetSoftwareVersion()
 
 	c := &controller{
-		db:        db,
+		state:     state,
 		fedDB:     federatingDB,
 		clock:     clock,
 		client:    client,
@@ -138,7 +138,7 @@ func (c *controller) NewTransportForUsername(ctx context.Context, username strin
 		u = username
 	}
 
-	ourAccount, err := c.db.GetAccountByUsernameDomain(ctx, u, "")
+	ourAccount, err := c.state.DB.GetAccountByUsernameDomain(ctx, u, "")
 	if err != nil {
 		return nil, fmt.Errorf("error getting account %s from db: %s", username, err)
 	}

internal/transport/finger.go

@@ -20,29 +20,61 @@ package transport
 
 import (
 	"context"
+	"encoding/xml"
 	"fmt"
 	"io"
 	"net/http"
+	"net/url"
 
+	apimodel "github.com/superseriousbusiness/gotosocial/internal/api/model"
 	apiutil "github.com/superseriousbusiness/gotosocial/internal/api/util"
 )
 
-func (t *transport) Finger(ctx context.Context, targetUsername string, targetDomain string) ([]byte, error) {
-	// Prepare URL string
-	urlStr := "https://" +
-		targetDomain +
-		"/.well-known/webfinger?resource=acct:" +
-		targetUsername + "@" + targetDomain
+// webfingerURLFor returns the URL to try a webfinger request against, as
+// well as if the URL was retrieved from cache. When the URL is retrieved
+// from cache we don't have to try and do host-meta discovery
+func (t *transport) webfingerURLFor(targetDomain string) (string, bool) {
+	url := "https://" + targetDomain + "/.well-known/webfinger"
 
-	// Generate new GET request from URL string
-	req, err := http.NewRequestWithContext(ctx, "GET", urlStr, nil)
+	wc := t.controller.state.Caches.GTS.Webfinger()
+	// We're doing the manual locking/unlocking here to be able to
+	// safely call Cache.Get instead of Get, as the latter updates the
+	// item expiry which we don't want to do here
+	wc.Lock()
+	item, ok := wc.Cache.Get(targetDomain)
+	wc.Unlock()
+
+	if ok {
+		url = item.Value
+	}
+
+	return url, ok
+}
+
+func prepWebfingerReq(ctx context.Context, loc, domain, username string) (*http.Request, error) {
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, loc, nil)
 	if err != nil {
 		return nil, err
 	}
+
+	value := url.QueryEscape("acct:" + username + "@" + domain)
+	req.URL.RawQuery = "resource=" + value
+
 	req.Header.Add("Accept", string(apiutil.AppJSON))
 	req.Header.Add("Accept", "application/jrd+json")
 	req.Header.Set("Host", req.URL.Host)
 
+	return req, nil
+}
+
+func (t *transport) Finger(ctx context.Context, targetUsername string, targetDomain string) ([]byte, error) {
+	// Generate new GET request
+	url, cached := t.webfingerURLFor(targetDomain)
+	req, err := prepWebfingerReq(ctx, url, targetDomain, targetUsername)
+	if err != nil {
+		return nil, err
+	}
+
 	// Perform the HTTP request
 	rsp, err := t.GET(req)
 	if err != nil {
@@ -50,10 +82,117 @@ func (t *transport) Finger(ctx context.Context, targetUsername string, targetDom
 	}
 	defer rsp.Body.Close()
 
-	// Check for an expected status code
-	if rsp.StatusCode != http.StatusOK {
-		return nil, fmt.Errorf("GET request to %s failed: %s", urlStr, rsp.Status)
+	// Check if the request succeeded so we can bail out early
+	if rsp.StatusCode == http.StatusOK {
+		if cached {
+			// If we got a success on a cached URL, i.e one set by us later on when
+			// a host-meta based webfinger request succeeded, set it again here to
+			// renew the TTL
+			t.controller.state.Caches.GTS.Webfinger().Set(targetDomain, url)
+		}
+		return io.ReadAll(rsp.Body)
 	}
 
+	// From here on out, we're handling different failure scenarios and
+	// deciding whether we should do a host-meta based fallback or not
+
+	if (rsp.StatusCode >= 500 && rsp.StatusCode < 600) || cached {
+		// In case we got a 5xx, bail out irrespective of if the value
+		// was cached or not. The target may be broken or be signalling
+		// us to back-off.
+		//
+		// If it's any error but the URL was cached, bail out too
+		return nil, fmt.Errorf("GET request to %s failed: %s", req.URL.String(), rsp.Status)
+	}
+
+	// So far we've failed to get a successful response from the expected
+	// webfinger endpoint. Lets try and discover the webfinger endpoint
+	// through /.well-known/host-meta
+	host, err := t.webfingerFromHostMeta(ctx, targetDomain)
+	if err != nil {
+		return nil, fmt.Errorf("failed to discover webfinger URL fallback for: %s through host-meta: %w", targetDomain, err)
+	}
+
+	// Check if the original and host-meta URL are the same. If they
+	// are there's no sense in us trying the request again as it just
+	// failed
+	if host == url {
+		return nil, fmt.Errorf("webfinger discovery on %s returned endpoint we already tried: %s", targetDomain, host)
+	}
+
+	// Now that we have a different URL for the webfinger
+	// endpoint, try the request against that endpoint instead
+	req, err = prepWebfingerReq(ctx, host, targetDomain, targetUsername)
+	if err != nil {
+		return nil, err
+	}
+
+	// Perform the HTTP request
+	rsp, err = t.GET(req)
+	if err != nil {
+		return nil, err
+	}
+	defer rsp.Body.Close()
+
+	if rsp.StatusCode != http.StatusOK {
+		// We've reached the end of the line here, both the original request
+		// and our attempt to resolve it through the fallback have failed
+		return nil, fmt.Errorf("GET request to %s failed: %s", req.URL.String(), rsp.Status)
+	}
+
+	// Set the URL in cache here, since host-meta told us this should be the
+	// valid one, it's different from the default and our request to it did
+	// not fail in any manner
+	t.controller.state.Caches.GTS.Webfinger().Set(targetDomain, host)
+
 	return io.ReadAll(rsp.Body)
 }
+
+func (t *transport) webfingerFromHostMeta(ctx context.Context, targetDomain string) (string, error) {
+	// Build the request for the host-meta endpoint
+	hmurl := "https://" + targetDomain + "/.well-known/host-meta"
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, hmurl, nil)
+	if err != nil {
+		return "", err
+	}
+
+	// We're doing XML
+	req.Header.Add("Accept", string(apiutil.AppXML))
+	req.Header.Add("Accept", "application/xrd+xml")
+	req.Header.Set("Host", req.URL.Host)
+
+	// Perform the HTTP request
+	rsp, err := t.GET(req)
+	if err != nil {
+		return "", err
+	}
+	defer rsp.Body.Close()
+
+	// Doesn't look like host-meta is working for this instance
+	if rsp.StatusCode != http.StatusOK {
+		return "", fmt.Errorf("GET request for %s failed: %s", req.URL.String(), rsp.Status)
+	}
+
+	e := xml.NewDecoder(rsp.Body)
+	var hm apimodel.HostMeta
+	if err := e.Decode(&hm); err != nil {
+		// We got something, but it's not a host-meta document we understand
+		return "", fmt.Errorf("failed to decode host-meta response for %s at %s: %w", targetDomain, req.URL.String(), err)
+	}
+
+	for _, link := range hm.Link {
+		// Based on what we currently understand, there should not be more than one
+		// of these with Rel="lrdd" in a host-meta document
+		if link.Rel == "lrdd" {
+			u, err := url.Parse(link.Template)
+			if err != nil {
+				return "", fmt.Errorf("lrdd link is not a valid url: %w", err)
+			}
+			// Get rid of the query template, we only want the scheme://host/path part
+			u.RawQuery = ""
+			urlStr := u.String()
+			return urlStr, nil
+		}
+	}
+	return "", fmt.Errorf("no webfinger URL found")
+}

internal/transport/finger_test.go

@@ -0,0 +1,118 @@
+/*
+   GoToSocial
+   Copyright (C) 2021-2023 GoToSocial Authors admin@gotosocial.org
+
+   This program is free software: you can redistribute it and/or modify
+   it under the terms of the GNU Affero General Public License as published by
+   the Free Software Foundation, either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU Affero General Public License for more details.
+
+   You should have received a copy of the GNU Affero General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+package transport_test
+
+import (
+	"context"
+	"testing"
+
+	"github.com/stretchr/testify/suite"
+)
+
+type FingerTestSuite struct {
+	TransportTestSuite
+}
+
+func (suite *FingerTestSuite) TestFinger() {
+	wc := suite.state.Caches.GTS.Webfinger()
+	suite.Equal(0, wc.Len(), "expect webfinger cache to be empty")
+
+	_, err := suite.transport.Finger(context.TODO(), "brand_new_person", "unknown-instance.com")
+	if err != nil {
+		suite.FailNow(err.Error())
+	}
+
+	suite.Equal(0, wc.Len(), "expect webfinger cache to be empty for normal webfinger request")
+}
+
+func (suite *FingerTestSuite) TestFingerWithHostMeta() {
+	wc := suite.state.Caches.GTS.Webfinger()
+	suite.Equal(0, wc.Len(), "expect webfinger cache to be empty")
+
+	_, err := suite.transport.Finger(context.TODO(), "someone", "misconfigured-instance.com")
+	if err != nil {
+		suite.FailNow(err.Error())
+	}
+
+	suite.Equal(1, wc.Len(), "expect webfinger cache to hold one entry")
+	suite.True(wc.Has("misconfigured-instance.com"), "expect webfinger cache to have entry for misconfigured-instance.com")
+}
+
+func (suite *FingerTestSuite) TestFingerWithHostMetaCacheStrategy() {
+	wc := suite.state.Caches.GTS.Webfinger()
+	suite.Equal(0, wc.Len(), "expect webfinger cache to be empty")
+
+	_, err := suite.transport.Finger(context.TODO(), "someone", "misconfigured-instance.com")
+	if err != nil {
+		suite.FailNow(err.Error())
+	}
+
+	suite.Equal(1, wc.Len(), "expect webfinger cache to hold one entry")
+	wc.Lock()
+	suite.True(wc.Cache.Has("misconfigured-instance.com"), "expect webfinger cache to have entry for misconfigured-instance.com")
+	ent, _ := wc.Cache.Get("misconfigured-instance.com")
+	wc.Unlock()
+
+	initialTime := ent.Expiry
+
+	// finger them again
+	_, err = suite.transport.Finger(context.TODO(), "someone", "misconfigured-instance.com")
+	if err != nil {
+		suite.FailNow(err.Error())
+	}
+
+	// there should still only be 1 cache entry
+	suite.Equal(1, wc.Len(), "expect webfinger cache to hold one entry")
+	wc.Lock()
+	suite.True(wc.Cache.Has("misconfigured-instance.com"), "expect webfinger cache to have entry for misconfigured-instance.com")
+	rep, _ := wc.Cache.Get("misconfigured-instance.com")
+	wc.Unlock()
+
+	repeatTime := rep.Expiry
+
+	// the TTL of the entry should have extended because we did a second
+	// successful finger
+	suite.NotEqual(initialTime, repeatTime, "expected webfinger cache entry to have different expiry times")
+	if repeatTime.Before(initialTime) {
+		suite.FailNow("expected webfinger cache entry to not be a time traveller")
+	}
+
+	// finger a non-existing user on that same instance which will return an error
+	_, err = suite.transport.Finger(context.TODO(), "invalid", "misconfigured-instance.com")
+	if err == nil {
+		suite.FailNow("expected request for invalid user to fail")
+	}
+
+	// there should still only be 1 cache entry, because we don't evict from cache on failure
+	suite.Equal(1, wc.Len(), "expect webfinger cache to hold one entry")
+	wc.Lock()
+	suite.True(wc.Cache.Has("misconfigured-instance.com"), "expect webfinger cache to have entry for misconfigured-instance.com")
+	last, _ := wc.Cache.Get("misconfigured-instance.com")
+	wc.Unlock()
+
+	lastTime := last.Expiry
+
+	// The TTL of the previous and new entry should be the same since
+	// a failed request must not extend the entry TTL
+	suite.Equal(repeatTime, lastTime)
+}
+
+func TestFingerTestSuite(t *testing.T) {
+	suite.Run(t, &FingerTestSuite{})
+}

internal/transport/transport_test.go

@@ -0,0 +1,101 @@
+/*
+   GoToSocial
+   Copyright (C) 2021-2023 GoToSocial Authors admin@gotosocial.org
+
+   This program is free software: you can redistribute it and/or modify
+   it under the terms of the GNU Affero General Public License as published by
+   the Free Software Foundation, either version 3 of the License, or
+   (at your option) any later version.
+
+   This program is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU Affero General Public License for more details.
+
+   You should have received a copy of the GNU Affero General Public License
+   along with this program.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+package transport_test
+
+import (
+	"context"
+
+	"github.com/stretchr/testify/suite"
+	"github.com/superseriousbusiness/gotosocial/internal/db"
+	"github.com/superseriousbusiness/gotosocial/internal/email"
+	"github.com/superseriousbusiness/gotosocial/internal/federation"
+	"github.com/superseriousbusiness/gotosocial/internal/gtsmodel"
+	"github.com/superseriousbusiness/gotosocial/internal/media"
+	"github.com/superseriousbusiness/gotosocial/internal/processing"
+	"github.com/superseriousbusiness/gotosocial/internal/state"
+	"github.com/superseriousbusiness/gotosocial/internal/storage"
+	"github.com/superseriousbusiness/gotosocial/internal/transport"
+	"github.com/superseriousbusiness/gotosocial/testrig"
+)
+
+type TransportTestSuite struct {
+	// standard suite interfaces
+	suite.Suite
+	db           db.DB
+	storage      *storage.Driver
+	mediaManager media.Manager
+	federator    federation.Federator
+	processor    *processing.Processor
+	emailSender  email.Sender
+	sentEmails   map[string]string
+	state        state.State
+
+	// standard suite models
+	testTokens       map[string]*gtsmodel.Token
+	testClients      map[string]*gtsmodel.Client
+	testApplications map[string]*gtsmodel.Application
+	testUsers        map[string]*gtsmodel.User
+	testAccounts     map[string]*gtsmodel.Account
+
+	transport transport.Transport
+}
+
+func (suite *TransportTestSuite) SetupSuite() {
+	suite.testTokens = testrig.NewTestTokens()
+	suite.testClients = testrig.NewTestClients()
+	suite.testApplications = testrig.NewTestApplications()
+	suite.testUsers = testrig.NewTestUsers()
+	suite.testAccounts = testrig.NewTestAccounts()
+}
+
+func (suite *TransportTestSuite) SetupTest() {
+	suite.state.Caches.Init()
+	testrig.StartWorkers(&suite.state)
+
+	testrig.InitTestConfig()
+	testrig.InitTestLog()
+
+	suite.db = testrig.NewTestDB(&suite.state)
+	suite.state.DB = suite.db
+	suite.storage = testrig.NewInMemoryStorage()
+	suite.state.Storage = suite.storage
+
+	suite.mediaManager = testrig.NewTestMediaManager(&suite.state)
+	suite.federator = testrig.NewTestFederator(&suite.state, testrig.NewTestTransportController(&suite.state, testrig.NewMockHTTPClient(nil, "../../testrig/media")), suite.mediaManager)
+	suite.sentEmails = make(map[string]string)
+	suite.emailSender = testrig.NewEmailSender("../../web/template/", suite.sentEmails)
+	suite.processor = testrig.NewTestProcessor(&suite.state, suite.federator, suite.emailSender, suite.mediaManager)
+
+	testrig.StandardDBSetup(suite.db, nil)
+	testrig.StandardStorageSetup(suite.storage, "../../testrig/media")
+
+	ts, err := suite.federator.TransportController().NewTransportForUsername(context.TODO(), "")
+	if err != nil {
+		suite.FailNow(err.Error())
+	}
+	suite.transport = ts
+
+	suite.NoError(suite.processor.Start())
+}
+
+func (suite *TransportTestSuite) TearDownTest() {
+	testrig.StandardDBTeardown(suite.db)
+	testrig.StandardStorageTeardown(suite.storage)
+	testrig.StopWorkers(&suite.state)
+}