Browse Source

Merge tag 'v4.1.3' into lets-bump-hometown-to-mastodon-4.2

Conflict resolution: only the streaming API is interesting. The
filtering for local statuses is done before the other filters even kick
in to be sure.
lets-bump-hometown-to-mastodon-4.2
nachtjasmin 2 years ago
parent
commit
cc9dc8d902
No known key found for this signature in database
  1. 48
      CHANGELOG.md
  2. 128
      Gemfile.lock
  3. 2
      app/controllers/backups_controller.rb
  4. 9
      app/controllers/well_known/webfinger_controller.rb
  5. 17
      app/lib/request.rb
  6. 2
      app/lib/scope_parser.rb
  7. 2
      app/lib/text_formatter.rb
  8. 6
      app/lib/vacuum/access_tokens_vacuum.rb
  9. 2
      app/models/concerns/attachmentable.rb
  10. 11
      app/models/form/account_batch.rb
  11. 105
      app/workers/scheduler/accounts_statuses_cleanup_scheduler.rb
  12. 3
      app/workers/scheduler/indexing_scheduler.rb
  13. 2
      config/initializers/content_security_policy.rb
  14. 2
      lib/mastodon/version.rb
  15. 5
      lib/public_file_server_middleware.rb
  16. 50
      spec/controllers/well_known/webfinger_controller_spec.rb
  17. 10
      spec/lib/vacuum/access_tokens_vacuum_spec.rb
  18. 63
      spec/models/form/account_batch_spec.rb
  19. 49
      spec/workers/scheduler/accounts_statuses_cleanup_scheduler_spec.rb
  20. 198
      streaming/index.js

48
CHANGELOG.md

@ -3,6 +3,54 @@ Changelog
All notable changes to this project will be documented in this file.
## [4.1.3] - 2023-07-06
### Added
- Add fallback redirection when getting a webfinger query `LOCAL_DOMAIN@LOCAL_DOMAIN` ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/23600))
### Changed
- Change OpenGraph-based embeds to allow fullscreen ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25058))
- Change AccessTokensVacuum to also delete expired tokens ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/24868))
- Change profile updates to be sent to recently-mentioned servers ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/24852))
- Change automatic post deletion thresholds and load detection ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/24614))
- Change `/api/v1/statuses/:id/history` to always return at least one item ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25510))
- Change auto-linking to allow carets in URL query params ([renchap](https://github.com/mastodon/mastodon/pull/25216))
### Removed
- Remove invalid `X-Frame-Options: ALLOWALL` ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25070))
### Fixed
- Fix wrong view being displayed when a webhook fails validation ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25464))
- Fix soft-deleted post cleanup scheduler overwhelming the streaming server ([ThisIsMissEm](https://github.com/mastodon/mastodon/pull/25519))
- Fix incorrect pagination headers in `/api/v2/admin/accounts` ([danielmbrasil](https://github.com/mastodon/mastodon/pull/25477))
- Fix multiple inefficiencies in automatic post cleanup worker ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/24607), [ClearlyClaire](https://github.com/mastodon/mastodon/pull/24785), [ClearlyClaire](https://github.com/mastodon/mastodon/pull/24840))
- Fix performance of streaming by parsing message JSON once ([ThisIsMissEm](https://github.com/mastodon/mastodon/pull/25278), [ThisIsMissEm](https://github.com/mastodon/mastodon/pull/25361))
- Fix CSP headers when `S3_ALIAS_HOST` includes a path component ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25273))
- Fix `tootctl accounts approve --number N` not aproving N earliest registrations ([danielmbrasil](https://github.com/mastodon/mastodon/pull/24605))
- Fix reports not being closed when performing batch suspensions ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/24988))
- Fix being able to vote on your own polls ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25015))
- Fix race condition when reblogging a status ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25016))
- Fix “Authorized applications” inefficiently and incorrectly getting last use date ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25060))
- Fix “Authorized applications” crashing when listing apps with certain admin API scopes ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25713))
- Fix multiple N+1s in ConversationsController ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25134), [ClearlyClaire](https://github.com/mastodon/mastodon/pull/25399), [ClearlyClaire](https://github.com/mastodon/mastodon/pull/25499))
- Fix user archive takeouts when using OpenStack Swift ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/24431))
- Fix searching for remote content by URL not working under certain conditions ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25637))
- Fix inefficiencies in indexing content for search ([VyrCossont](https://github.com/mastodon/mastodon/pull/24285), [VyrCossont](https://github.com/mastodon/mastodon/pull/24342))
### Security
- Add finer permission requirements for managing webhooks ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25463))
- Update dependencies
- Add hardening headers for user-uploaded files ([ClearlyClaire](https://github.com/mastodon/mastodon/pull/25756))
- Fix verified links possibly hiding important parts of the URL (CVE-2023-36462)
- Fix timeout handling of outbound HTTP requests (CVE-2023-36461)
- Fix arbitrary file creation through media processing (CVE-2023-36460)
- Fix possible XSS in preview cards (CVE-2023-36459)
## [4.1.2] - 2023-04-04
### Fixed

128
Gemfile.lock

@ -10,40 +10,40 @@ GIT
GEM
remote: https://rubygems.org/
specs:
actioncable (6.1.7.2)
actionpack (= 6.1.7.2)
activesupport (= 6.1.7.2)
actioncable (6.1.7.4)
actionpack (= 6.1.7.4)
activesupport (= 6.1.7.4)
nio4r (~> 2.0)
websocket-driver (>= 0.6.1)
actionmailbox (6.1.7.2)
actionpack (= 6.1.7.2)
activejob (= 6.1.7.2)
activerecord (= 6.1.7.2)
activestorage (= 6.1.7.2)
activesupport (= 6.1.7.2)
actionmailbox (6.1.7.4)
actionpack (= 6.1.7.4)
activejob (= 6.1.7.4)
activerecord (= 6.1.7.4)
activestorage (= 6.1.7.4)
activesupport (= 6.1.7.4)
mail (>= 2.7.1)
actionmailer (6.1.7.2)
actionpack (= 6.1.7.2)
actionview (= 6.1.7.2)
activejob (= 6.1.7.2)
activesupport (= 6.1.7.2)
actionmailer (6.1.7.4)
actionpack (= 6.1.7.4)
actionview (= 6.1.7.4)
activejob (= 6.1.7.4)
activesupport (= 6.1.7.4)
mail (~> 2.5, >= 2.5.4)
rails-dom-testing (~> 2.0)
actionpack (6.1.7.2)
actionview (= 6.1.7.2)
activesupport (= 6.1.7.2)
actionpack (6.1.7.4)
actionview (= 6.1.7.4)
activesupport (= 6.1.7.4)
rack (~> 2.0, >= 2.0.9)
rack-test (>= 0.6.3)
rails-dom-testing (~> 2.0)
rails-html-sanitizer (~> 1.0, >= 1.2.0)
actiontext (6.1.7.2)
actionpack (= 6.1.7.2)
activerecord (= 6.1.7.2)
activestorage (= 6.1.7.2)
activesupport (= 6.1.7.2)
actiontext (6.1.7.4)
actionpack (= 6.1.7.4)
activerecord (= 6.1.7.4)
activestorage (= 6.1.7.4)
activesupport (= 6.1.7.4)
nokogiri (>= 1.8.5)
actionview (6.1.7.2)
activesupport (= 6.1.7.2)
actionview (6.1.7.4)
activesupport (= 6.1.7.4)
builder (~> 3.1)
erubi (~> 1.4)
rails-dom-testing (~> 2.0)
@ -54,22 +54,22 @@ GEM
case_transform (>= 0.2)
jsonapi-renderer (>= 0.1.1.beta1, < 0.3)
active_record_query_trace (1.8)
activejob (6.1.7.2)
activesupport (= 6.1.7.2)
activejob (6.1.7.4)
activesupport (= 6.1.7.4)
globalid (>= 0.3.6)
activemodel (6.1.7.2)
activesupport (= 6.1.7.2)
activerecord (6.1.7.2)
activemodel (= 6.1.7.2)
activesupport (= 6.1.7.2)
activestorage (6.1.7.2)
actionpack (= 6.1.7.2)
activejob (= 6.1.7.2)
activerecord (= 6.1.7.2)
activesupport (= 6.1.7.2)
activemodel (6.1.7.4)
activesupport (= 6.1.7.4)
activerecord (6.1.7.4)
activemodel (= 6.1.7.4)
activesupport (= 6.1.7.4)
activestorage (6.1.7.4)
actionpack (= 6.1.7.4)
activejob (= 6.1.7.4)
activerecord (= 6.1.7.4)
activesupport (= 6.1.7.4)
marcel (~> 1.0)
mini_mime (>= 1.1.0)
activesupport (6.1.7.2)
activesupport (6.1.7.4)
concurrent-ruby (~> 1.0, >= 1.0.2)
i18n (>= 1.6, < 2)
minitest (>= 5.1)
@ -173,7 +173,7 @@ GEM
cocoon (1.2.15)
coderay (1.1.3)
color_diff (0.1)
concurrent-ruby (1.2.0)
concurrent-ruby (1.2.2)
connection_pool (2.3.0)
cose (1.2.1)
cbor (~> 0.5.9)
@ -206,7 +206,7 @@ GEM
docile (1.4.0)
domain_name (0.5.20190701)
unf (>= 0.0.5, < 1.0.0)
doorkeeper (5.6.4)
doorkeeper (5.6.6)
railties (>= 5)
dotenv (2.8.1)
dotenv-rails (2.8.1)
@ -388,7 +388,7 @@ GEM
loofah (2.19.1)
crass (~> 1.0.2)
nokogiri (>= 1.5.9)
mail (2.8.0.1)
mail (2.8.1)
mini_mime (>= 0.1.1)
net-imap
net-pop
@ -405,12 +405,12 @@ GEM
mime-types-data (~> 3.2015)
mime-types-data (3.2022.0105)
mini_mime (1.1.2)
mini_portile2 (2.8.1)
mini_portile2 (2.8.2)
minitest (5.17.0)
msgpack (1.6.0)
multi_json (1.15.0)
multipart-post (2.1.1)
net-imap (0.3.4)
net-imap (0.3.6)
date
net-protocol
net-ldap (0.17.1)
@ -423,8 +423,8 @@ GEM
net-smtp (0.3.3)
net-protocol
net-ssh (7.0.1)
nio4r (2.5.8)
nokogiri (1.14.1)
nio4r (2.5.9)
nokogiri (1.14.5)
mini_portile2 (~> 2.8.0)
racc (~> 1.4)
nsa (0.2.8)
@ -497,7 +497,7 @@ GEM
activesupport (>= 3.0.0)
raabro (1.4.0)
racc (1.6.2)
rack (2.2.6.2)
rack (2.2.7)
rack-attack (6.6.1)
rack (>= 1.0, < 3)
rack-cors (1.1.1)
@ -512,20 +512,20 @@ GEM
rack
rack-test (2.0.2)
rack (>= 1.3)
rails (6.1.7.2)
actioncable (= 6.1.7.2)
actionmailbox (= 6.1.7.2)
actionmailer (= 6.1.7.2)
actionpack (= 6.1.7.2)
actiontext (= 6.1.7.2)
actionview (= 6.1.7.2)
activejob (= 6.1.7.2)
activemodel (= 6.1.7.2)
activerecord (= 6.1.7.2)
activestorage (= 6.1.7.2)
activesupport (= 6.1.7.2)
rails (6.1.7.4)
actioncable (= 6.1.7.4)
actionmailbox (= 6.1.7.4)
actionmailer (= 6.1.7.4)
actionpack (= 6.1.7.4)
actiontext (= 6.1.7.4)
actionview (= 6.1.7.4)
activejob (= 6.1.7.4)
activemodel (= 6.1.7.4)
activerecord (= 6.1.7.4)
activestorage (= 6.1.7.4)
activesupport (= 6.1.7.4)
bundler (>= 1.15.0)
railties (= 6.1.7.2)
railties (= 6.1.7.4)
sprockets-rails (>= 2.0.0)
rails-controller-testing (1.0.5)
actionpack (>= 5.0.1.rc1)
@ -541,9 +541,9 @@ GEM
railties (>= 6.0.0, < 7)
rails-settings-cached (0.6.6)
rails (>= 4.2.0)
railties (6.1.7.2)
actionpack (= 6.1.7.2)
activesupport (= 6.1.7.2)
railties (6.1.7.4)
actionpack (= 6.1.7.4)
activesupport (= 6.1.7.4)
method_source
rake (>= 12.2)
thor (~> 1.0)
@ -688,9 +688,9 @@ GEM
unicode-display_width (>= 1.1.1, < 3)
terrapin (0.6.0)
climate_control (>= 0.0.3, < 1.0)
thor (1.2.1)
thor (1.2.2)
tilt (2.0.11)
timeout (0.3.1)
timeout (0.3.2)
tpm-key_attestation (0.11.0)
bindata (~> 2.4)
openssl (> 2.0, < 3.1)
@ -753,7 +753,7 @@ GEM
xorcist (1.1.3)
xpath (3.2.0)
nokogiri (~> 1.8)
zeitwerk (2.6.6)
zeitwerk (2.6.8)
PLATFORMS
ruby

2
app/controllers/backups_controller.rb

@ -13,7 +13,7 @@ class BackupsController < ApplicationController
when :s3
redirect_to @backup.dump.expiring_url(10)
when :fog
if Paperclip::Attachment.default_options.dig(:storage, :fog_credentials, :openstack_temp_url_key).present?
if Paperclip::Attachment.default_options.dig(:fog_credentials, :openstack_temp_url_key).present?
redirect_to @backup.dump.expiring_url(Time.now.utc + 10)
else
redirect_to full_asset_url(@backup.dump.url)

9
app/controllers/well_known/webfinger_controller.rb

@ -18,7 +18,14 @@ module WellKnown
private
def set_account
@account = Account.find_local!(username_from_resource)
username = username_from_resource
@account = begin
if username == Rails.configuration.x.local_domain
Account.representative
else
Account.find_local!(username)
end
end
end
def username_from_resource

17
app/lib/request.rb

@ -11,15 +11,7 @@ require 'resolv'
# Also changes how the read timeout behaves so that it is cumulative (closer
# to HTTP::Timeout::Global, but still having distinct timeouts for other
# operation types)
class PerOperationWithDeadline < HTTP::Timeout::PerOperation
READ_DEADLINE = 30
def initialize(*args)
super
@read_deadline = options.fetch(:read_deadline, READ_DEADLINE)
end
class HTTP::Timeout::PerOperation
def connect(socket_class, host, port, nodelay = false)
@socket = socket_class.open(host, port)
@socket.setsockopt(Socket::IPPROTO_TCP, Socket::TCP_NODELAY, 1) if nodelay
@ -32,7 +24,7 @@ class PerOperationWithDeadline < HTTP::Timeout::PerOperation
# Read data from the socket
def readpartial(size, buffer = nil)
@deadline ||= Process.clock_gettime(Process::CLOCK_MONOTONIC) + @read_deadline
@deadline ||= Process.clock_gettime(Process::CLOCK_MONOTONIC) + @read_timeout
timeout = false
loop do
@ -41,8 +33,7 @@ class PerOperationWithDeadline < HTTP::Timeout::PerOperation
return :eof if result.nil?
remaining_time = @deadline - Process.clock_gettime(Process::CLOCK_MONOTONIC)
raise HTTP::TimeoutError, "Read timed out after #{@read_timeout} seconds" if timeout
raise HTTP::TimeoutError, "Read timed out after a total of #{@read_deadline} seconds" if remaining_time <= 0
raise HTTP::TimeoutError, "Read timed out after #{@read_timeout} seconds" if timeout || remaining_time <= 0
return result if result != :wait_readable
# marking the socket for timeout. Why is this not being raised immediately?
@ -55,7 +46,7 @@ class PerOperationWithDeadline < HTTP::Timeout::PerOperation
# timeout. Else, the first timeout was a proper timeout.
# This hack has to be done because io/wait#wait_readable doesn't provide a value for when
# the socket is closed by the server, and HTTP::Parser doesn't provide the limit for the chunks.
timeout = true unless @socket.to_io.wait_readable([remaining_time, @read_timeout].min)
timeout = true unless @socket.to_io.wait_readable(remaining_time)
end
end
end

2
app/lib/scope_parser.rb

@ -1,7 +1,7 @@
# frozen_string_literal: true
class ScopeParser < Parslet::Parser
rule(:term) { match('[a-z]').repeat(1).as(:term) }
rule(:term) { match('[a-z_]').repeat(1).as(:term) }
rule(:colon) { str(':') }
rule(:access) { (str('write') | str('read')).as(:access) }
rule(:namespace) { str('admin').as(:namespace) }

2
app/lib/text_formatter.rb

@ -60,7 +60,7 @@ class TextFormatter
suffix = url[prefix.length + 30..-1]
cutoff = url[prefix.length..-1].length > 30
<<~HTML.squish.html_safe # rubocop:disable Rails/OutputSafety
<<~HTML.squish
<a href="#{h(url)}" target="_blank" rel="#{rel.join(' ')}"><span class="invisible">#{h(prefix)}</span><span class="#{cutoff ? 'ellipsis' : ''}">#{h(display_url)}</span><span class="invisible">#{h(suffix)}</span></a>
HTML
rescue Addressable::URI::InvalidURIError, IDN::Idna::IdnaError

6
app/lib/vacuum/access_tokens_vacuum.rb

@ -9,10 +9,12 @@ class Vacuum::AccessTokensVacuum
private
def vacuum_revoked_access_tokens!
Doorkeeper::AccessToken.where.not(revoked_at: nil).where('revoked_at < NOW()').delete_all
Doorkeeper::AccessToken.where.not(expires_in: nil).where('created_at + make_interval(secs => expires_in) < NOW()').in_batches.delete_all
Doorkeeper::AccessToken.where.not(revoked_at: nil).where('revoked_at < NOW()').in_batches.delete_all
end
def vacuum_revoked_access_grants!
Doorkeeper::AccessGrant.where.not(revoked_at: nil).where('revoked_at < NOW()').delete_all
Doorkeeper::AccessGrant.where.not(expires_in: nil).where('created_at + make_interval(secs => expires_in) < NOW()').in_batches.delete_all
Doorkeeper::AccessGrant.where.not(revoked_at: nil).where('revoked_at < NOW()').in_batches.delete_all
end
end

2
app/models/concerns/attachmentable.rb

@ -24,7 +24,7 @@ module Attachmentable
def self.has_attached_file(name, options = {}) # rubocop:disable Naming/PredicateName
super(name, options)
send(:"before_#{name}_validate", prepend: true) do
send(:"before_#{name}_validate") do
attachment = send(name)
check_image_dimension(attachment)
set_file_content_type(attachment)

11
app/models/form/account_batch.rb

@ -123,7 +123,18 @@ class Form::AccountBatch
account: current_account,
action: :suspend
)
Admin::SuspensionWorker.perform_async(account.id)
# Suspending a single account closes their associated reports, so
# mass-suspending would be consistent.
Report.where(target_account: account).unresolved.find_each do |report|
authorize(report, :update?)
log_action(:resolve, report)
report.resolve!(current_account)
rescue Mastodon::NotPermittedError
# This should not happen, but just in case, do not fail early
end
end
def approve_account(account)

105
app/workers/scheduler/accounts_statuses_cleanup_scheduler.rb

@ -7,28 +7,30 @@ class Scheduler::AccountsStatusesCleanupScheduler
# This limit is mostly to be nice to the fediverse at large and not
# generate too much traffic.
# This also helps limiting the running time of the scheduler itself.
MAX_BUDGET = 150
MAX_BUDGET = 300
# This is an attempt to spread the load across instances, as various
# accounts are likely to have various followers.
# This is an attempt to spread the load across remote servers, as
# spreading deletions across diverse accounts is likely to spread
# the deletion across diverse followers. It also helps each individual
# user see some effect sooner.
PER_ACCOUNT_BUDGET = 5
# This is an attempt to limit the workload generated by status removal
# jobs to something the particular instance can handle.
PER_THREAD_BUDGET = 6
# Those avoid loading an instance that is already under load
MAX_DEFAULT_SIZE = 200
MAX_DEFAULT_LATENCY = 5
MAX_PUSH_SIZE = 500
MAX_PUSH_LATENCY = 10
# 'pull' queue has lower priority jobs, and it's unlikely that pushing
# deletes would cause much issues with this queue if it didn't cause issues
# with default and push. Yet, do not enqueue deletes if the instance is
# lagging behind too much.
MAX_PULL_SIZE = 10_000
MAX_PULL_LATENCY = 5.minutes.to_i
# jobs to something the particular server can handle.
PER_THREAD_BUDGET = 5
# These are latency limits on various queues above which a server is
# considered to be under load, causing the auto-deletion to be entirely
# skipped for that run.
LOAD_LATENCY_THRESHOLDS = {
default: 5,
push: 10,
# The `pull` queue has lower priority jobs, and it's unlikely that
# pushing deletes would cause much issues with this queue if it didn't
# cause issues with `default` and `push`. Yet, do not enqueue deletes
# if the instance is lagging behind too much.
pull: 5.minutes.to_i,
}.freeze
sidekiq_options retry: 0, lock: :until_executed, lock_ttl: 1.day.to_i
@ -36,17 +38,37 @@ class Scheduler::AccountsStatusesCleanupScheduler
return if under_load?
budget = compute_budget
first_policy_id = last_processed_id
# If the budget allows it, we want to consider all accounts with enabled
# auto cleanup at least once.
#
# We start from `first_policy_id` (the last processed id in the previous
# run) and process each policy until we loop to `first_policy_id`,
# recording into `affected_policies` any policy that caused posts to be
# deleted.
#
# After that, we set `full_iteration` to `false` and continue looping on
# policies from `affected_policies`.
first_policy_id = last_processed_id || 0
first_iteration = true
full_iteration = true
affected_policies = []
loop do
num_processed_accounts = 0
scope = AccountStatusesCleanupPolicy.where(enabled: true)
scope.where(Account.arel_table[:id].gt(first_policy_id)) if first_policy_id.present?
scope = cleanup_policies(first_policy_id, affected_policies, first_iteration, full_iteration)
scope.find_each(order: :asc) do |policy|
num_deleted = AccountStatusesCleanupService.new.call(policy, [budget, PER_ACCOUNT_BUDGET].min)
num_processed_accounts += 1 unless num_deleted.zero?
budget -= num_deleted
unless num_deleted.zero?
num_processed_accounts += 1
affected_policies << policy.id if full_iteration
end
full_iteration = false if !first_iteration && policy.id >= first_policy_id
if budget.zero?
save_last_processed_id(policy.id)
break
@ -55,36 +77,55 @@ class Scheduler::AccountsStatusesCleanupScheduler
# The idea here is to loop through all policies at least once until the budget is exhausted
# and start back after the last processed account otherwise
break if budget.zero? || (num_processed_accounts.zero? && first_policy_id.nil?)
first_policy_id = nil
break if budget.zero? || (num_processed_accounts.zero? && !full_iteration)
full_iteration = false unless first_iteration
first_iteration = false
end
end
def compute_budget
threads = Sidekiq::ProcessSet.new.select { |x| x['queues'].include?('push') }.map { |x| x['concurrency'] }.sum
# Each post deletion is a `RemovalWorker` job (on `default` queue), each
# potentially spawning many `ActivityPub::DeliveryWorker` jobs (on the `push` queue).
threads = Sidekiq::ProcessSet.new.select { |x| x['queues'].include?('push') }.pluck('concurrency').sum
[PER_THREAD_BUDGET * threads, MAX_BUDGET].min
end
def under_load?
queue_under_load?('default', MAX_DEFAULT_SIZE, MAX_DEFAULT_LATENCY) || queue_under_load?('push', MAX_PUSH_SIZE, MAX_PUSH_LATENCY) || queue_under_load?('pull', MAX_PULL_SIZE, MAX_PULL_LATENCY)
LOAD_LATENCY_THRESHOLDS.any? { |queue, max_latency| queue_under_load?(queue, max_latency) }
end
private
def queue_under_load?(name, max_size, max_latency)
queue = Sidekiq::Queue.new(name)
queue.size > max_size || queue.latency > max_latency
def cleanup_policies(first_policy_id, affected_policies, first_iteration, full_iteration)
scope = AccountStatusesCleanupPolicy.where(enabled: true)
if full_iteration
# If we are doing a full iteration, examine all policies we have not examined yet
if first_iteration
scope.where(id: first_policy_id...)
else
scope.where(id: ..first_policy_id).or(scope.where(id: affected_policies))
end
else
# Otherwise, examine only policies that previously yielded posts to delete
scope.where(id: affected_policies)
end
end
def queue_under_load?(name, max_latency)
Sidekiq::Queue.new(name).latency > max_latency
end
def last_processed_id
redis.get('account_statuses_cleanup_scheduler:last_account_id')
redis.get('account_statuses_cleanup_scheduler:last_policy_id')&.to_i
end
def save_last_processed_id(id)
if id.nil?
redis.del('account_statuses_cleanup_scheduler:last_account_id')
redis.del('account_statuses_cleanup_scheduler:last_policy_id')
else
redis.set('account_statuses_cleanup_scheduler:last_account_id', id, ex: 1.hour.seconds)
redis.set('account_statuses_cleanup_scheduler:last_policy_id', id, ex: 1.hour.seconds)
end
end
end

3
app/workers/scheduler/indexing_scheduler.rb

@ -9,6 +9,9 @@ class Scheduler::IndexingScheduler
IMPORT_BATCH_SIZE = 1000
SCAN_BATCH_SIZE = 10 * IMPORT_BATCH_SIZE
IMPORT_BATCH_SIZE = 1000
SCAN_BATCH_SIZE = 10 * IMPORT_BATCH_SIZE
def perform
return unless Chewy.enabled?

2
config/initializers/content_security_policy.rb

@ -3,7 +3,7 @@
# https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Content-Security-Policy
def host_to_url(str)
"http#{Rails.configuration.x.use_https ? 's' : ''}://#{str.split('/').first}" if str.present?
"http#{Rails.configuration.x.use_https ? 's' : ''}://#{str}".split('/').first if str.present?
end
base_host = Rails.configuration.x.web_domain

2
lib/mastodon/version.rb

@ -13,7 +13,7 @@ module Mastodon
end
def patch
2
3
end
def flags

5
lib/public_file_server_middleware.rb

@ -32,6 +32,11 @@ class PublicFileServerMiddleware
end
end
# Override the default CSP header set by the CSP middleware
headers['Content-Security-Policy'] = "default-src 'none'; form-action 'none'" if request_path.start_with?(paperclip_root_url)
headers['X-Content-Type-Options'] = 'nosniff'
[status, headers, response]
end

50
spec/controllers/well_known/webfinger_controller_spec.rb

@ -4,6 +4,10 @@ describe WellKnown::WebfingerController, type: :controller do
render_views
describe 'GET #show' do
subject(:perform_show!) do
get :show, params: { resource: resource }, format: :json
end
let(:alternate_domains) { [] }
let(:alice) { Fabricate(:account, username: 'alice') }
let(:resource) { nil }
@ -15,10 +19,6 @@ describe WellKnown::WebfingerController, type: :controller do
Rails.configuration.x.alternate_domains = tmp
end
subject do
get :show, params: { resource: resource }, format: :json
end
shared_examples 'a successful response' do
it 'returns http success' do
expect(response).to have_http_status(200)
@ -43,7 +43,7 @@ describe WellKnown::WebfingerController, type: :controller do
let(:resource) { alice.to_webfinger_s }
before do
subject
perform_show!
end
it_behaves_like 'a successful response'
@ -54,7 +54,7 @@ describe WellKnown::WebfingerController, type: :controller do
before do
alice.suspend!
subject
perform_show!
end
it_behaves_like 'a successful response'
@ -66,7 +66,7 @@ describe WellKnown::WebfingerController, type: :controller do
before do
alice.suspend!
alice.deletion_request.destroy
subject
perform_show!
end
it 'returns http gone' do
@ -78,7 +78,7 @@ describe WellKnown::WebfingerController, type: :controller do
let(:resource) { 'acct:not@existing.com' }
before do
subject
perform_show!
end
it 'returns http not found' do
@ -90,7 +90,7 @@ describe WellKnown::WebfingerController, type: :controller do
let(:alternate_domains) { ['foo.org'] }
before do
subject
perform_show!
end
context 'when an account exists' do
@ -114,11 +114,39 @@ describe WellKnown::WebfingerController, type: :controller do
end
end
context 'when the old name scheme is used to query the instance actor' do
let(:resource) do
"#{Rails.configuration.x.local_domain}@#{Rails.configuration.x.local_domain}"
end
before do
perform_show!
end
it 'returns http success' do
expect(response).to have_http_status(200)
end
it 'does not set a Vary header' do
expect(response.headers['Vary']).to be_nil
end
it 'returns application/jrd+json' do
expect(response.media_type).to eq 'application/jrd+json'
end
it 'returns links for the internal account' do
json = body_as_json
expect(json[:subject]).to eq 'acct:mastodon.internal@cb6e6126.ngrok.io'
expect(json[:aliases]).to eq ['https://cb6e6126.ngrok.io/actor']
end
end
context 'with no resource parameter' do
let(:resource) { nil }
before do
subject
perform_show!
end
it 'returns http bad request' do
@ -130,7 +158,7 @@ describe WellKnown::WebfingerController, type: :controller do
let(:resource) { 'df/:dfkj' }
before do
subject
perform_show!
end
it 'returns http bad request' do

10
spec/lib/vacuum/access_tokens_vacuum_spec.rb

@ -5,9 +5,11 @@ RSpec.describe Vacuum::AccessTokensVacuum do
describe '#perform' do
let!(:revoked_access_token) { Fabricate(:access_token, revoked_at: 1.minute.ago) }
let!(:expired_access_token) { Fabricate(:access_token, expires_in: 59.minutes.to_i, created_at: 1.hour.ago) }
let!(:active_access_token) { Fabricate(:access_token) }
let!(:revoked_access_grant) { Fabricate(:access_grant, revoked_at: 1.minute.ago) }
let!(:expired_access_grant) { Fabricate(:access_grant, expires_in: 59.minutes.to_i, created_at: 1.hour.ago) }
let!(:active_access_grant) { Fabricate(:access_grant) }
before do
@ -18,10 +20,18 @@ RSpec.describe Vacuum::AccessTokensVacuum do
expect { revoked_access_token.reload }.to raise_error ActiveRecord::RecordNotFound
end
it 'deletes expired access tokens' do
expect { expired_access_token.reload }.to raise_error ActiveRecord::RecordNotFound
end
it 'deletes revoked access grants' do
expect { revoked_access_grant.reload }.to raise_error ActiveRecord::RecordNotFound
end
it 'deletes expired access grants' do
expect { expired_access_grant.reload }.to raise_error ActiveRecord::RecordNotFound
end
it 'does not delete active access tokens' do
expect { active_access_token.reload }.to_not raise_error
end

63
spec/models/form/account_batch_spec.rb

@ -0,0 +1,63 @@
# frozen_string_literal: true
require 'rails_helper'
RSpec.describe Form::AccountBatch do
let(:account_batch) { described_class.new }
describe '#save' do
subject { account_batch.save }
let(:account) { Fabricate(:user, role: UserRole.find_by(name: 'Admin')).account }
let(:account_ids) { [] }
let(:query) { Account.none }
before do
account_batch.assign_attributes(
action: action,
current_account: account,
account_ids: account_ids,
query: query,
select_all_matching: select_all_matching
)
end
context 'when action is "suspend"' do
let(:action) { 'suspend' }
let(:target_account) { Fabricate(:account) }
let(:target_account2) { Fabricate(:account) }
before do
Fabricate(:report, target_account: target_account)
Fabricate(:report, target_account: target_account2)
end
context 'when accounts are passed as account_ids' do
let(:select_all_matching) { '0' }
let(:account_ids) { [target_account.id, target_account2.id] }
it 'suspends the expected users' do
expect { subject }.to change { [target_account.reload.suspended?, target_account2.reload.suspended?] }.from([false, false]).to([true, true])
end
it 'closes open reports targeting the suspended users' do
expect { subject }.to change { Report.unresolved.where(target_account: [target_account, target_account2]).count }.from(2).to(0)
end
end
context 'when accounts are passed as a query' do
let(:select_all_matching) { '1' }
let(:query) { Account.where(id: [target_account.id, target_account2.id]) }
it 'suspends the expected users' do
expect { subject }.to change { [target_account.reload.suspended?, target_account2.reload.suspended?] }.from([false, false]).to([true, true])
end
it 'closes open reports targeting the suspended users' do
expect { subject }.to change { Report.unresolved.where(target_account: [target_account, target_account2]).count }.from(2).to(0)
end
end
end
end
end

49
spec/workers/scheduler/accounts_statuses_cleanup_scheduler_spec.rb

@ -7,11 +7,13 @@ describe Scheduler::AccountsStatusesCleanupScheduler do
let!(:account2) { Fabricate(:account, domain: nil) }
let!(:account3) { Fabricate(:account, domain: nil) }
let!(:account4) { Fabricate(:account, domain: nil) }
let!(:account5) { Fabricate(:account, domain: nil) }
let!(:remote) { Fabricate(:account) }
let!(:policy1) { Fabricate(:account_statuses_cleanup_policy, account: account1) }
let!(:policy2) { Fabricate(:account_statuses_cleanup_policy, account: account3) }
let!(:policy3) { Fabricate(:account_statuses_cleanup_policy, account: account4, enabled: false) }
let!(:policy4) { Fabricate(:account_statuses_cleanup_policy, account: account5) }
let(:queue_size) { 0 }
let(:queue_latency) { 0 }
@ -40,6 +42,7 @@ describe Scheduler::AccountsStatusesCleanupScheduler do
Fabricate(:status, account: account2, created_at: 3.years.ago)
Fabricate(:status, account: account3, created_at: 3.years.ago)
Fabricate(:status, account: account4, created_at: 3.years.ago)
Fabricate(:status, account: account5, created_at: 3.years.ago)
Fabricate(:status, account: remote, created_at: 3.years.ago)
end
@ -70,7 +73,7 @@ describe Scheduler::AccountsStatusesCleanupScheduler do
end
end
describe '#get_budget' do
describe '#compute_budget' do
context 'on a single thread' do
let(:process_set_stub) { [ { 'concurrency' => 1, 'queues' => ['push', 'default'] } ] }
@ -109,8 +112,48 @@ describe Scheduler::AccountsStatusesCleanupScheduler do
expect { subject.perform }.to_not change { account4.statuses.count }
end
it 'eventually deletes every deletable toot' do
expect { subject.perform; subject.perform; subject.perform; subject.perform }.to change { Status.count }.by(-20)
it 'eventually deletes every deletable toot given enough runs' do
stub_const 'Scheduler::AccountsStatusesCleanupScheduler::MAX_BUDGET', 4
expect { 10.times { subject.perform } }.to change { Status.count }.by(-30)
end
it 'correctly round-trips between users across several runs' do
stub_const 'Scheduler::AccountsStatusesCleanupScheduler::MAX_BUDGET', 3
stub_const 'Scheduler::AccountsStatusesCleanupScheduler::PER_ACCOUNT_BUDGET', 2
expect { 3.times { subject.perform } }
.to change { Status.count }.by(-3 * 3)
.and change { account1.statuses.count }
.and change { account3.statuses.count }
.and change { account5.statuses.count }
end
context 'when given a big budget' do
let(:process_set_stub) { [{ 'concurrency' => 400, 'queues' => %w(push default) }] }
before do
stub_const 'Scheduler::AccountsStatusesCleanupScheduler::MAX_BUDGET', 400
end
it 'correctly handles looping in a single run' do
expect(subject.compute_budget).to eq(400)
expect { subject.perform }.to change { Status.count }.by(-30)
end
end
context 'when there is no work to be done' do
let(:process_set_stub) { [{ 'concurrency' => 400, 'queues' => %w(push default) }] }
before do
stub_const 'Scheduler::AccountsStatusesCleanupScheduler::MAX_BUDGET', 400
subject.perform
end
it 'does not get stuck' do
expect(subject.compute_budget).to eq(400)
expect { subject.perform }.to_not change { Status.count }
end
end
end
end

198
streaming/index.js

@ -226,15 +226,9 @@ const startWorker = async (workerId) => {
callbacks.forEach(callback => callback(json));
};
/**
* @callback SubscriptionListener
* @param {ReturnType<parseJSON>} json of the message
* @returns void
*/
/**
* @param {string} channel
* @param {SubscriptionListener} callback
* @param {function(string): void} callback
*/
const subscribe = (channel, callback) => {
log.silly(`Adding listener for ${channel}`);
@ -251,7 +245,7 @@ const startWorker = async (workerId) => {
/**
* @param {string} channel
* @param {SubscriptionListener} callback
* @param {function(Object<string, any>): void} callback
*/
const unsubscribe = (channel, callback) => {
log.silly(`Removing listener for ${channel}`);
@ -629,29 +623,27 @@ const startWorker = async (workerId) => {
* @param {string[]} ids
* @param {any} req
* @param {function(string, string): void} output
* @param {undefined | function(string[], SubscriptionListener): void} attachCloseHandler
* @param {function(string[], function(string): void): void} attachCloseHandler
* @param {boolean=} needsFiltering
* @returns {SubscriptionListener}
* @returns {function(object): void}
*/
const streamFrom = (ids, req, output, attachCloseHandler, needsFiltering = false) => {
const accountId = req.accountId || req.remoteAddress;
log.verbose(req.requestId, `Starting stream from ${ids.join(', ')} for ${accountId}`);
const transmit = (event, payload) => {
// TODO: Replace "string"-based delete payloads with object payloads:
const encodedPayload = typeof payload === 'object' ? JSON.stringify(payload) : payload;
// Currently message is of type string, soon it'll be Record<string, any>
const listener = message => {
const { event, payload, queued_at } = message;
log.silly(req.requestId, `Transmitting for ${accountId}: ${event} ${encodedPayload}`);
output(event, encodedPayload);
};
const transmit = () => {
const now = new Date().getTime();
const delta = now - queued_at;
const encodedPayload = typeof payload === 'object' ? JSON.stringify(payload) : payload;
// The listener used to process each message off the redis subscription,
// message here is an object with an `event` and `payload` property. Some
// events also include a queued_at value, but this is being removed shortly.
/** @type {SubscriptionListener} */
const listener = message => {
const { event, payload } = message;
log.silly(req.requestId, `Transmitting for ${accountId}: ${event} ${encodedPayload} Delay: ${delta}ms`);
output(event, encodedPayload);
};
// Only send local-only statuses to logged-in users
if (payload.local_only && !req.accountId) {
@ -659,42 +651,29 @@ const startWorker = async (workerId) => {
return;
}
// Streaming only needs to apply filtering to some channels and only to
// some events. This is because majority of the filtering happens on the
// Ruby on Rails side when producing the event for streaming.
//
// The only events that require filtering from the streaming server are
// `update` and `status.update`, all other events are transmitted to the
// client as soon as they're received (pass-through).
//
// The channels that need filtering are determined in the function
// `channelNameToIds` defined below:
if (!needsFiltering || (event !== 'update' && event !== 'status.update')) {
transmit(event, payload);
// Only messages that may require filtering are statuses, since notifications
// are already personalized and deletes do not matter
if (!needsFiltering || event !== 'update') {
transmit();
return;
}
// The rest of the logic from here on in this function is to handle
// filtering of statuses:
const unpackedPayload = payload;
const targetAccountIds = [unpackedPayload.account.id].concat(unpackedPayload.mentions.map(item => item.id));
const accountDomain = unpackedPayload.account.acct.split('@')[1];
// Filter based on language:
if (Array.isArray(req.chosenLanguages) && payload.language !== null && req.chosenLanguages.indexOf(payload.language) === -1) {
log.silly(req.requestId, `Message ${payload.id} filtered by language (${payload.language})`);
if (Array.isArray(req.chosenLanguages) && unpackedPayload.language !== null && req.chosenLanguages.indexOf(unpackedPayload.language) === -1) {
log.silly(req.requestId, `Message ${unpackedPayload.id} filtered by language (${unpackedPayload.language})`);
return;
}
// When the account is not logged in, it is not necessary to confirm the block or mute
if (!req.accountId) {
transmit(event, payload);
transmit();
return;
}
// Filter based on domain blocks, blocks, mutes, or custom filters:
const targetAccountIds = [payload.account.id].concat(payload.mentions.map(item => item.id));
const accountDomain = payload.account.acct.split('@')[1];
// TODO: Move this logic out of the message handling loop
pgPool.connect((err, client, releasePgConnection) => {
pgPool.connect((err, client, done) => {
if (err) {
log.error(err);
return;
@ -709,57 +688,40 @@ const startWorker = async (workerId) => {
SELECT 1
FROM mutes
WHERE account_id = $1
AND target_account_id IN (${placeholders(targetAccountIds, 2)})`, [req.accountId, payload.account.id].concat(targetAccountIds)),
AND target_account_id IN (${placeholders(targetAccountIds, 2)})`, [req.accountId, unpackedPayload.account.id].concat(targetAccountIds)),
];
if (accountDomain) {
queries.push(client.query('SELECT 1 FROM account_domain_blocks WHERE account_id = $1 AND domain = $2', [req.accountId, accountDomain]));
}
if (!payload.filtered && !req.cachedFilters) {
if (!unpackedPayload.filtered && !req.cachedFilters) {
queries.push(client.query('SELECT filter.id AS id, filter.phrase AS title, filter.context AS context, filter.expires_at AS expires_at, filter.action AS filter_action, keyword.keyword AS keyword, keyword.whole_word AS whole_word FROM custom_filter_keywords keyword JOIN custom_filters filter ON keyword.custom_filter_id = filter.id WHERE filter.account_id = $1 AND (filter.expires_at IS NULL OR filter.expires_at > NOW())', [req.accountId]));
}
Promise.all(queries).then(values => {
releasePgConnection();
done();
// Handling blocks & mutes and domain blocks: If one of those applies,
// then we don't transmit the payload of the event to the client
if (values[0].rows.length > 0 || (accountDomain && values[1].rows.length > 0)) {
return;
}
// If the payload already contains the `filtered` property, it means
// that filtering has been applied on the ruby on rails side, as
// such, we don't need to construct or apply the filters in streaming:
if (Object.prototype.hasOwnProperty.call(payload, "filtered")) {
transmit(event, payload);
return;
}
// Handling for constructing the custom filters and caching them on the request
// TODO: Move this logic out of the message handling lifecycle
if (!req.cachedFilters) {
if (!unpackedPayload.filtered && !req.cachedFilters) {
const filterRows = values[accountDomain ? 2 : 1].rows;
req.cachedFilters = filterRows.reduce((cache, filter) => {
if (cache[filter.id]) {
cache[filter.id].keywords.push([filter.keyword, filter.whole_word]);
req.cachedFilters = filterRows.reduce((cache, row) => {
if (cache[row.id]) {
cache[row.id].keywords.push([row.keyword, row.whole_word]);
} else {
cache[filter.id] = {
keywords: [[filter.keyword, filter.whole_word]],
expires_at: filter.expires_at,
filter: {
id: filter.id,
title: filter.title,
context: filter.context,
expires_at: filter.expires_at,
// filter.filter_action is the value from the
// custom_filters.action database column, it is an integer
// representing a value in an enum defined by Ruby on Rails:
//
// enum { warn: 0, hide: 1 }
filter_action: ['warn', 'hide'][filter.filter_action],
cache[row.id] = {
keywords: [[row.keyword, row.whole_word]],
expires_at: row.expires_at,
repr: {
id: row.id,
title: row.title,
context: row.context,
expires_at: row.expires_at,
filter_action: ['warn', 'hide'][row.filter_action],
},
};
}
@ -767,10 +729,6 @@ const startWorker = async (workerId) => {
return cache;
}, {});
// Construct the regular expressions for the custom filters: This
// needs to be done in a separate loop as the database returns one
// filterRow per keyword, so we need all the keywords before
// constructing the regular expression
Object.keys(req.cachedFilters).forEach((key) => {
req.cachedFilters[key].regexp = new RegExp(req.cachedFilters[key].keywords.map(([keyword, whole_word]) => {
let expr = keyword.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
@ -790,58 +748,31 @@ const startWorker = async (workerId) => {
});
}
// Apply cachedFilters against the payload, constructing a
// `filter_results` array of FilterResult entities
if (req.cachedFilters) {
const status = payload;
// TODO: Calculate searchableContent in Ruby on Rails:
const searchableContent = ([status.spoiler_text || '', status.content].concat((status.poll && status.poll.options) ? status.poll.options.map(option => option.title) : [])).concat(status.media_attachments.map(att => att.description)).join('\n\n').replace(/<br\s*\/?>/g, '\n').replace(/<\/p><p>/g, '\n\n');
const searchableTextContent = JSDOM.fragment(searchableContent).textContent;
// Check filters
if (req.cachedFilters && !unpackedPayload.filtered) {
const status = unpackedPayload;
const searchContent = ([status.spoiler_text || '', status.content].concat((status.poll && status.poll.options) ? status.poll.options.map(option => option.title) : [])).concat(status.media_attachments.map(att => att.description)).join('\n\n').replace(/<br\s*\/?>/g, '\n').replace(/<\/p><p>/g, '\n\n');
const searchIndex = JSDOM.fragment(searchContent).textContent;
const now = new Date();
const filter_results = Object.values(req.cachedFilters).reduce((results, cachedFilter) => {
// Check the filter hasn't expired before applying:
if (cachedFilter.expires_at !== null && cachedFilter.expires_at < now) {
return results;
}
// Just in-case JSDOM fails to find textContent in searchableContent
if (!searchableTextContent) {
return results;
}
const keyword_matches = searchableTextContent.match(cachedFilter.regexp);
if (keyword_matches) {
// results is an Array of FilterResult; status_matches is always
// null as we only are only applying the keyword-based custom
// filters, not the status-based custom filters.
// https://docs.joinmastodon.org/entities/FilterResult/
results.push({
filter: cachedFilter.filter,
keyword_matches,
status_matches: null
});
payload.filtered = [];
Object.values(req.cachedFilters).forEach((cachedFilter) => {
if ((cachedFilter.expires_at === null || cachedFilter.expires_at > now)) {
const keyword_matches = searchIndex.match(cachedFilter.regexp);
if (keyword_matches) {
payload.filtered.push({
filter: cachedFilter.repr,
keyword_matches,
});
}
}
return results;
}, []);
// Send the payload + the FilterResults as the `filtered` property
// to the streaming connection. To reach this code, the `event` must
// have been either `update` or `status.update`, meaning the
// `payload` is a Status entity, which has a `filtered` property:
//
// filtered: https://docs.joinmastodon.org/entities/Status/#filtered
transmit(event, {
...payload,
filtered: filter_results
});
} else {
transmit(event, payload);
}
transmit();
}).catch(err => {
releasePgConnection();
log.error(err);
done();
});
});
};
@ -850,7 +781,7 @@ const startWorker = async (workerId) => {
subscribe(`${redisPrefix}${id}`, listener);
});
if (typeof attachCloseHandler === 'function') {
if (attachCloseHandler) {
attachCloseHandler(ids.map(id => `${redisPrefix}${id}`), listener);
}
@ -887,13 +818,12 @@ const startWorker = async (workerId) => {
/**
* @param {any} req
* @param {function(): void} [closeHandler]
* @returns {function(string[], SubscriptionListener): void}
* @return {function(string[]): void}
*/
const streamHttpEnd = (req, closeHandler = undefined) => (ids, listener) => {
const streamHttpEnd = (req, closeHandler = undefined) => (ids) => {
req.on('close', () => {
ids.forEach(id => {
unsubscribe(id, listener);
unsubscribe(id);
});
if (closeHandler) {
@ -1153,7 +1083,7 @@ const startWorker = async (workerId) => {
* @typedef WebSocketSession
* @property {any} socket
* @property {any} request
* @property {Object.<string, { listener: SubscriptionListener, stopHeartbeat: function(): void }>} subscriptions
* @property {Object.<string, { listener: function(string): void, stopHeartbeat: function(): void }>} subscriptions
*/
/**

Loading…
Cancel
Save