Merge commit from fork

* Add limit on inbox payload size

The 1MB limit is consistent with the limit we use when fetching remote resources

* Add limit to number of options from federated polls

* Add a limit to the number of federated profile fields

* Add limit on federated username length

* Add hard limits for federated display name and account bio

* Add hard limits for `alsoKnownAs` and `attributionDomains`

* Add hard limit on federated custom emoji shortcode

* Highlight most destructive limits and expand on their reasoning

FEDERATION.md

@@ -48,3 +48,22 @@ Mastodon requires all `POST` requests to be signed, and MAY require `GET` reques
 ### Additional documentation
 
 - [Mastodon documentation](https://docs.joinmastodon.org/)
+
+## Size limits
+
+Mastodon imposes a few hard limits on federated content.
+These limits are intended to be very generous and way above what the Mastodon user experience is optimized for, so as to accomodate future changes and unusual or unforeseen usage patterns, while still providing some limits for performance reasons.
+The following table attempts to summary those limits.
+
+| Limited property                                              | Size limit | Consequence of exceeding the limit |
+| ------------------------------------------------------------- | ---------- | ---------------------------------- |
+| Serialized JSON-LD                                            | 1MB        | **Activity is rejected/dropped**   |
+| Profile fields (actor `PropertyValue` attachments) name/value | 2047       | Field name/value is truncated      |
+| Number of profile fields (actor `PropertyValue` attachments)  | 50         | Fields list is truncated           |
+| Poll options (number of `anyOf`/`oneOf` in a `Question`)      | 500        | Items list is truncated            |
+| Account username (actor `preferredUsername`) length           | 2048       | **Actor will be rejected**         |
+| Account display name (actor `name`) length                    | 2048       | Display name will be truncated     |
+| Account note (actor `summary`) length                         | 20kB       | Account note will be truncated     |
+| Account `attributionDomains`                                  | 256        | List will be truncated             |
+| Account aliases (actor `alsoKnownAs`)                         | 256        | List will be truncated             |
+| Custom emoji shortcode (`Emoji` `name`)                       | 2048       | Emoji will be rejected             |

app/controllers/activitypub/inboxes_controller.rb

@@ -3,6 +3,7 @@
 class ActivityPub::InboxesController < ActivityPub::BaseController
   include JsonLdHelper
 
+  before_action :skip_large_payload
   before_action :skip_unknown_actor_activity
   before_action :require_actor_signature!
   skip_before_action :authenticate_user!
@@ -16,6 +17,10 @@ class ActivityPub::InboxesController < ActivityPub::BaseController
 
   private
 
+  def skip_large_payload
+    head 413 if request.content_length > ActivityPub::Activity::MAX_JSON_SIZE
+  end
+
   def skip_unknown_actor_activity
     head 202 if unknown_affected_account?
   end

app/lib/activitypub/activity.rb

@@ -5,6 +5,7 @@ class ActivityPub::Activity
   include Redisable
   include Lockable
 
+  MAX_JSON_SIZE = 1.megabyte
   SUPPORTED_TYPES = %w(Note Question).freeze
   CONVERTED_TYPES = %w(Image Audio Video Article Page Event).freeze
 

app/lib/activitypub/parser/poll_parser.rb

@@ -3,6 +3,10 @@
 class ActivityPub::Parser::PollParser
   include JsonLdHelper
 
+  # Limit the number of items for performance purposes.
+  # We truncate rather than error out to avoid missing the post entirely.
+  MAX_ITEMS = 500
+
   def initialize(json)
     @json = json
   end
@@ -48,6 +52,6 @@ class ActivityPub::Parser::PollParser
   private
 
   def items
-    @json['anyOf'] || @json['oneOf']
+    (@json['anyOf'] || @json['oneOf'])&.take(MAX_ITEMS)
   end
 end

app/models/account.rb

@@ -80,6 +80,13 @@ class Account < ApplicationRecord
   DISPLAY_NAME_LENGTH_LIMIT = 30
   NOTE_LENGTH_LIMIT = 500
 
+  # Hard limits for federated content
+  USERNAME_LENGTH_HARD_LIMIT = 2048
+  DISPLAY_NAME_LENGTH_HARD_LIMIT = 2048
+  NOTE_LENGTH_HARD_LIMIT = 20.kilobytes
+  ATTRIBUTION_DOMAINS_HARD_LIMIT = 256
+  ALSO_KNOWN_AS_HARD_LIMIT = 256
+
   AUTOMATED_ACTOR_TYPES = %w(Application Service).freeze
 
   include Attachmentable # Load prior to Avatar & Header concerns
@@ -112,7 +119,7 @@ class Account < ApplicationRecord
   validates_with UniqueUsernameValidator, if: -> { will_save_change_to_username? }
 
   # Remote user validations, also applies to internal actors
-  validates :username, format: { with: USERNAME_ONLY_RE }, if: -> { (remote? || actor_type_application?) && will_save_change_to_username? }
+  validates :username, format: { with: USERNAME_ONLY_RE }, length: { maximum: USERNAME_LENGTH_HARD_LIMIT }, if: -> { (remote? || actor_type_application?) && will_save_change_to_username? }
 
   # Remote user validations
   validates :uri, presence: true, unless: :local?, on: :create

app/models/custom_emoji.rb

@@ -26,6 +26,8 @@ class CustomEmoji < ApplicationRecord
 
   LIMIT = 256.kilobytes
   MINIMUM_SHORTCODE_SIZE = 2
+  MAX_SHORTCODE_SIZE = 128
+  MAX_FEDERATED_SHORTCODE_SIZE = 2048
 
   SHORTCODE_RE_FRAGMENT = '[a-zA-Z0-9_]{2,}'
 
@@ -45,7 +47,8 @@ class CustomEmoji < ApplicationRecord
   normalizes :domain, with: ->(domain) { domain.downcase.strip }
 
   validates_attachment :image, content_type: { content_type: IMAGE_MIME_TYPES }, presence: true, size: { less_than: LIMIT }
-  validates :shortcode, uniqueness: { scope: :domain }, format: { with: SHORTCODE_ONLY_RE }, length: { minimum: MINIMUM_SHORTCODE_SIZE }
+  validates :shortcode, uniqueness: { scope: :domain }, format: { with: SHORTCODE_ONLY_RE }, length: { minimum: MINIMUM_SHORTCODE_SIZE, maximum: MAX_FEDERATED_SHORTCODE_SIZE }
+  validates :shortcode, length: { maximum: MAX_SHORTCODE_SIZE }, if: :local?
 
   scope :local, -> { where(domain: nil) }
   scope :remote, -> { where.not(domain: nil) }

app/services/activitypub/process_account_service.rb

@@ -6,6 +6,7 @@ class ActivityPub::ProcessAccountService < BaseService
   include Redisable
   include Lockable
 
+  MAX_PROFILE_FIELDS = 50
   SUBDOMAINS_RATELIMIT = 10
   DISCOVERIES_PER_REQUEST = 400
 
@@ -123,15 +124,15 @@ class ActivityPub::ProcessAccountService < BaseService
 
   def set_immediate_attributes!
     @account.featured_collection_url = valid_collection_uri(@json['featured'])
-    @account.display_name            = @json['name'] || ''
-    @account.note                    = @json['summary'] || ''
+    @account.display_name            = (@json['name'] || '')[0...(Account::DISPLAY_NAME_LENGTH_HARD_LIMIT)]
+    @account.note                    = (@json['summary'] || '')[0...(Account::NOTE_LENGTH_HARD_LIMIT)]
     @account.locked                  = @json['manuallyApprovesFollowers'] || false
     @account.fields                  = property_values || {}
-    @account.also_known_as           = as_array(@json['alsoKnownAs'] || []).map { |item| value_or_id(item) }
+    @account.also_known_as           = as_array(@json['alsoKnownAs'] || []).take(Account::ALSO_KNOWN_AS_HARD_LIMIT).map { |item| value_or_id(item) }
     @account.discoverable            = @json['discoverable'] || false
     @account.indexable               = @json['indexable'] || false
     @account.memorial                = @json['memorial'] || false
-    @account.attribution_domains     = as_array(@json['attributionDomains'] || []).map { |item| value_or_id(item) }
+    @account.attribution_domains     = as_array(@json['attributionDomains'] || []).take(Account::ATTRIBUTION_DOMAINS_HARD_LIMIT).map { |item| value_or_id(item) }
   end
 
   def set_fetchable_key!
@@ -252,7 +253,10 @@ class ActivityPub::ProcessAccountService < BaseService
   def property_values
     return unless @json['attachment'].is_a?(Array)
 
-    as_array(@json['attachment']).select { |attachment| attachment['type'] == 'PropertyValue' }.map { |attachment| attachment.slice('name', 'value') }
+    as_array(@json['attachment'])
+      .select { |attachment| attachment['type'] == 'PropertyValue' }
+      .take(MAX_PROFILE_FIELDS)
+      .map { |attachment| attachment.slice('name', 'value') }
   end
 
   def mismatching_origin?(url)