diff --git a/app/models/changeset.rb b/app/models/changeset.rb index abb494de6..e6bde19a5 100644 --- a/app/models/changeset.rb +++ b/app/models/changeset.rb @@ -130,6 +130,8 @@ class Changeset < ApplicationRecord def update_bbox!(bbox_update) bbox.expand!(bbox_update) + raise OSM::APISizeLimitExceeded if bbox.linear_size > size_limit + # update active record. rails 2.1's dirty handling should take care of # whether this object needs saving or not. self.min_lon, self.min_lat, self.max_lon, self.max_lat = @bbox.to_a.collect(&:round) if bbox.complete? @@ -225,4 +227,10 @@ class Changeset < ApplicationRecord def subscribed?(user) subscribers.exists?(user.id) end + + def size_limit + @size_limit ||= ActiveRecord::Base.connection.select_value( + "SELECT api_size_limit($1)", "api_size_limit", [user_id] + ) + end end diff --git a/config/settings.yml b/config/settings.yml index ec868b651..fa7207721 100644 --- a/config/settings.yml +++ b/config/settings.yml @@ -73,6 +73,13 @@ max_changes_per_hour: 100000 days_to_max_changes: 7 importer_changes_per_hour: 1000000 moderator_changes_per_hour: 1000000 +# Size limit for changes +min_size_limit: 10000000 +initial_size_limit: 30000000 +max_size_limit: 5400000000 +days_to_max_size_limit: 28 +importer_size_limit: 5400000000 +moderator_size_limit: 5400000000 # Domain for handling message replies #messages_domain: "messages.openstreetmap.org" # MaxMind GeoIPv2 database diff --git a/db/migrate/20240618193051_api_size_limit.rb b/db/migrate/20240618193051_api_size_limit.rb new file mode 100644 index 000000000..ce97a9f0e --- /dev/null +++ b/db/migrate/20240618193051_api_size_limit.rb @@ -0,0 +1,13 @@ +class ApiSizeLimit < ActiveRecord::Migration[7.1] + def up + safety_assured do + execute DatabaseFunctions::API_SIZE_LIMIT + end + end + + def down + safety_assured do + execute "DROP FUNCTION api_size_limit(bigint)" + end + end +end diff --git a/db/structure.sql b/db/structure.sql index 2e8bdbb61..e9d5efc3f 100644 --- a/db/structure.sql +++ b/db/structure.sql @@ -176,6 +176,63 @@ CREATE FUNCTION public.api_rate_limit(user_id bigint) RETURNS integer $$; +-- +-- Name: api_size_limit(bigint); Type: FUNCTION; Schema: public; Owner: - +-- + +CREATE FUNCTION public.api_size_limit(user_id bigint) RETURNS bigint + LANGUAGE plpgsql STABLE + AS $$ + DECLARE + min_size_limit int8 := 10000000; + initial_size_limit int8 := 30000000; + max_size_limit int8 := 5400000000; + days_to_max_size_limit int4 := 28; + importer_size_limit int8 := 5400000000; + moderator_size_limit int8 := 5400000000; + roles text[]; + last_block timestamp without time zone; + first_change timestamp without time zone; + active_reports int4; + time_since_first_change double precision; + size_limit int8; + BEGIN + SELECT ARRAY_AGG(user_roles.role) INTO STRICT roles FROM user_roles WHERE user_roles.user_id = api_size_limit.user_id; + + IF 'moderator' = ANY(roles) THEN + size_limit := moderator_size_limit; + ELSIF 'importer' = ANY(roles) THEN + size_limit := importer_size_limit; + ELSE + SELECT user_blocks.created_at INTO last_block FROM user_blocks WHERE user_blocks.user_id = api_size_limit.user_id ORDER BY user_blocks.created_at DESC LIMIT 1; + + IF FOUND THEN + SELECT changesets.created_at INTO first_change FROM changesets WHERE changesets.user_id = api_size_limit.user_id AND changesets.created_at > last_block ORDER BY changesets.created_at LIMIT 1; + ELSE + SELECT changesets.created_at INTO first_change FROM changesets WHERE changesets.user_id = api_size_limit.user_id ORDER BY changesets.created_at LIMIT 1; + END IF; + + IF NOT FOUND THEN + first_change := CURRENT_TIMESTAMP AT TIME ZONE 'UTC'; + END IF; + + SELECT COUNT(*) INTO STRICT active_reports + FROM issues INNER JOIN reports ON reports.issue_id = issues.id + WHERE issues.reported_user_id = api_size_limit.user_id AND issues.status = 'open' AND reports.updated_at >= COALESCE(issues.resolved_at, '1970-01-01'); + + time_since_first_change := EXTRACT(EPOCH FROM CURRENT_TIMESTAMP AT TIME ZONE 'UTC' - first_change); + + size_limit := max_size_limit * POWER(time_since_first_change, 2) / POWER(days_to_max_size_limit * 24 * 60 * 60, 2); + size_limit := GREATEST(initial_size_limit, LEAST(max_size_limit, FLOOR(size_limit))); + size_limit := size_limit / POWER(2, active_reports); + size_limit := GREATEST(min_size_limit, LEAST(max_size_limit, size_limit)); + END IF; + + RETURN size_limit; + END; + $$; + + SET default_tablespace = ''; SET default_table_access_method = heap; @@ -3521,6 +3578,7 @@ INSERT INTO "schema_migrations" (version) VALUES ('23'), ('22'), ('21'), +('20240618193051'), ('20240605134916'), ('20240405083825'), ('20240307181018'), diff --git a/lib/bounding_box.rb b/lib/bounding_box.rb index 0cc4c5fd4..462f45a9f 100644 --- a/lib/bounding_box.rb +++ b/lib/bounding_box.rb @@ -88,6 +88,14 @@ class BoundingBox end end + def linear_size + if complete? + (max_lon - min_lon) + (max_lat - min_lat) + else + 0 + end + end + def complete? to_a.exclude?(nil) end diff --git a/lib/database_functions.rb b/lib/database_functions.rb index f9e09ac70..d58736fdf 100644 --- a/lib/database_functions.rb +++ b/lib/database_functions.rb @@ -55,4 +55,58 @@ module DatabaseFunctions END; $$ LANGUAGE plpgsql STABLE; ).freeze + + API_SIZE_LIMIT = %( + CREATE OR REPLACE FUNCTION api_size_limit(user_id int8) + RETURNS int8 + AS $$ + DECLARE + min_size_limit int8 := #{Settings.min_size_limit}; + initial_size_limit int8 := #{Settings.initial_size_limit}; + max_size_limit int8 := #{Settings.max_size_limit}; + days_to_max_size_limit int4 := #{Settings.days_to_max_size_limit}; + importer_size_limit int8 := #{Settings.importer_size_limit}; + moderator_size_limit int8 := #{Settings.moderator_size_limit}; + roles text[]; + last_block timestamp without time zone; + first_change timestamp without time zone; + active_reports int4; + time_since_first_change double precision; + size_limit int8; + BEGIN + SELECT ARRAY_AGG(user_roles.role) INTO STRICT roles FROM user_roles WHERE user_roles.user_id = api_size_limit.user_id; + + IF 'moderator' = ANY(roles) THEN + size_limit := moderator_size_limit; + ELSIF 'importer' = ANY(roles) THEN + size_limit := importer_size_limit; + ELSE + SELECT user_blocks.created_at INTO last_block FROM user_blocks WHERE user_blocks.user_id = api_size_limit.user_id ORDER BY user_blocks.created_at DESC LIMIT 1; + + IF FOUND THEN + SELECT changesets.created_at INTO first_change FROM changesets WHERE changesets.user_id = api_size_limit.user_id AND changesets.created_at > last_block ORDER BY changesets.created_at LIMIT 1; + ELSE + SELECT changesets.created_at INTO first_change FROM changesets WHERE changesets.user_id = api_size_limit.user_id ORDER BY changesets.created_at LIMIT 1; + END IF; + + IF NOT FOUND THEN + first_change := CURRENT_TIMESTAMP AT TIME ZONE 'UTC'; + END IF; + + SELECT COUNT(*) INTO STRICT active_reports + FROM issues INNER JOIN reports ON reports.issue_id = issues.id + WHERE issues.reported_user_id = api_size_limit.user_id AND issues.status = 'open' AND reports.updated_at >= COALESCE(issues.resolved_at, '1970-01-01'); + + time_since_first_change := EXTRACT(EPOCH FROM CURRENT_TIMESTAMP AT TIME ZONE 'UTC' - first_change); + + size_limit := max_size_limit * POWER(time_since_first_change, 2) / POWER(days_to_max_size_limit * 24 * 60 * 60, 2); + size_limit := GREATEST(initial_size_limit, LEAST(max_size_limit, FLOOR(size_limit))); + size_limit := size_limit / POWER(2, active_reports); + size_limit := GREATEST(min_size_limit, LEAST(max_size_limit, size_limit)); + END IF; + + RETURN size_limit; + END; + $$ LANGUAGE plpgsql STABLE; + ).freeze end diff --git a/lib/osm.rb b/lib/osm.rb index a0fcef8b9..dd273418e 100644 --- a/lib/osm.rb +++ b/lib/osm.rb @@ -364,6 +364,17 @@ module OSM end end + # Raised when a size limit is exceeded + class APISizeLimitExceeded < APIError + def initialize + super("Size limit exceeded") + end + + def status + :payload_too_large + end + end + # Helper methods for going to/from mercator and lat/lng. class Mercator include Math diff --git a/lib/tasks/update_functions.rake b/lib/tasks/update_functions.rake index 605d3c9ad..f10257215 100644 --- a/lib/tasks/update_functions.rake +++ b/lib/tasks/update_functions.rake @@ -2,5 +2,6 @@ namespace :db do desc "Update database function definitions" task :update_functions => :environment do ActiveRecord::Base.connection.execute DatabaseFunctions::API_RATE_LIMIT + ActiveRecord::Base.connection.execute DatabaseFunctions::API_SIZE_LIMIT end end diff --git a/test/controllers/api/changesets_controller_test.rb b/test/controllers/api/changesets_controller_test.rb index be1033ea6..17a9ad9d3 100644 --- a/test/controllers/api/changesets_controller_test.rb +++ b/test/controllers/api/changesets_controller_test.rb @@ -749,7 +749,11 @@ module Api end def test_upload_large_changeset - auth_header = basic_authorization_header create(:user).email, "test" + user = create(:user) + auth_header = basic_authorization_header user.email, "test" + + # create an old changeset to ensure we have the maximum rate limit + create(:changeset, :user => user, :created_at => Time.now.utc - 28.days) # create a changeset put changeset_create_path, :params => "", :headers => auth_header @@ -1048,14 +1052,14 @@ module Api diff = <<~CHANGESET - - - - - - - - + + + + + + + + CHANGESET @@ -1329,9 +1333,9 @@ module Api diff = <<~CHANGESET - - - + + + @@ -1352,9 +1356,9 @@ module Api diff = <<~CHANGESET - - - + + + @@ -1384,9 +1388,9 @@ module Api diff = <<~CHANGESET - - - + + + @@ -1407,9 +1411,9 @@ module Api diff = <<~CHANGESET - - - + + + @@ -1478,14 +1482,14 @@ module Api changeset_id = @response.body.to_i old_way = create(:way) - create(:way_node, :way => old_way, :node => create(:node, :lat => 1, :lon => 1)) + create(:way_node, :way => old_way, :node => create(:node, :lat => 0.1, :lon => 0.1)) diff = XML::Document.new diff.root = XML::Node.new "osmChange" modify = XML::Node.new "modify" xml_old_way = xml_node_for_way(old_way) nd_ref = XML::Node.new "nd" - nd_ref["ref"] = create(:node, :lat => 3, :lon => 3).id.to_s + nd_ref["ref"] = create(:node, :lat => 0.3, :lon => 0.3).id.to_s xml_old_way << nd_ref xml_old_way["changeset"] = changeset_id.to_s modify << xml_old_way @@ -1498,10 +1502,10 @@ module Api # check the bbox changeset = Changeset.find(changeset_id) - assert_equal 1 * GeoRecord::SCALE, changeset.min_lon, "min_lon should be 1 degree" - assert_equal 3 * GeoRecord::SCALE, changeset.max_lon, "max_lon should be 3 degrees" - assert_equal 1 * GeoRecord::SCALE, changeset.min_lat, "min_lat should be 1 degree" - assert_equal 3 * GeoRecord::SCALE, changeset.max_lat, "max_lat should be 3 degrees" + assert_equal 0.1 * GeoRecord::SCALE, changeset.min_lon, "min_lon should be 0.1 degree" + assert_equal 0.3 * GeoRecord::SCALE, changeset.max_lon, "max_lon should be 0.3 degrees" + assert_equal 0.1 * GeoRecord::SCALE, changeset.min_lat, "min_lat should be 0.1 degree" + assert_equal 0.3 * GeoRecord::SCALE, changeset.max_lat, "max_lat should be 0.3 degrees" end ## @@ -1798,6 +1802,71 @@ module Api assert_response :too_many_requests, "upload did not hit rate limit" end + ## + # test initial size limit + def test_upload_initial_size_limit + # create a user + user = create(:user) + + # create a changeset that puts us near the initial size limit + changeset = create(:changeset, :user => user, + :min_lat => (-0.5 * GeoRecord::SCALE).round, :min_lon => (0.5 * GeoRecord::SCALE).round, + :max_lat => (0.5 * GeoRecord::SCALE).round, :max_lon => (2.5 * GeoRecord::SCALE).round) + + # create authentication header + auth_header = basic_authorization_header user.email, "test" + + # simple diff to create a node + diff = <<~CHANGESET + + + + + + + + + CHANGESET + + # upload it + post changeset_upload_path(changeset), :params => diff, :headers => auth_header + assert_response :payload_too_large, "upload did not hit size limit" + end + + ## + # test size limit after one week + def test_upload_week_size_limit + # create a user + user = create(:user) + + # create a changeset to establish our initial edit time + create(:changeset, :user => user, :created_at => Time.now.utc - 7.days) + + # create a changeset that puts us near the initial size limit + changeset = create(:changeset, :user => user, + :min_lat => (-0.5 * GeoRecord::SCALE).round, :min_lon => (0.5 * GeoRecord::SCALE).round, + :max_lat => (0.5 * GeoRecord::SCALE).round, :max_lon => (2.5 * GeoRecord::SCALE).round) + + # create authentication header + auth_header = basic_authorization_header user.email, "test" + + # simple diff to create a node way and relation using placeholders + diff = <<~CHANGESET + + + + + + + + + CHANGESET + + # upload it + post changeset_upload_path(changeset), :params => diff, :headers => auth_header + assert_response :payload_too_large, "upload did not hit size limit" + end + ## # when we make some simple changes we get the same changes back from the # diff download. @@ -1829,14 +1898,14 @@ module Api diff = <<~CHANGESET - - - - - - - - + + + + + + + + CHANGESET @@ -1935,15 +2004,15 @@ module Api diff = <<~CHANGESET - + - - - + + + - + @@ -2034,7 +2103,7 @@ module Api # FIXME: This should really be moded to a integration test due to the with_controller def test_changeset_bbox way = create(:way) - create(:way_node, :way => way, :node => create(:node, :lat => 3, :lon => 3)) + create(:way_node, :way => way, :node => create(:node, :lat => 0.3, :lon => 0.3)) auth_header = basic_authorization_header create(:user).email, "test" @@ -2046,7 +2115,7 @@ module Api # add a single node to it with_controller(NodesController.new) do - xml = "" + xml = "" put node_create_path, :params => xml, :headers => auth_header assert_response :success, "Couldn't create node." end @@ -2054,14 +2123,14 @@ module Api # get the bounding box back from the changeset get changeset_show_path(:id => changeset_id) assert_response :success, "Couldn't read back changeset." - assert_select "osm>changeset[min_lon='1.0000000']", 1 - assert_select "osm>changeset[max_lon='1.0000000']", 1 - assert_select "osm>changeset[min_lat='2.0000000']", 1 - assert_select "osm>changeset[max_lat='2.0000000']", 1 + assert_select "osm>changeset[min_lon='0.1000000']", 1 + assert_select "osm>changeset[max_lon='0.1000000']", 1 + assert_select "osm>changeset[min_lat='0.2000000']", 1 + assert_select "osm>changeset[max_lat='0.2000000']", 1 # add another node to it with_controller(NodesController.new) do - xml = "" + xml = "" put node_create_path, :params => xml, :headers => auth_header assert_response :success, "Couldn't create second node." end @@ -2069,10 +2138,10 @@ module Api # get the bounding box back from the changeset get changeset_show_path(:id => changeset_id) assert_response :success, "Couldn't read back changeset for the second time." - assert_select "osm>changeset[min_lon='1.0000000']", 1 - assert_select "osm>changeset[max_lon='2.0000000']", 1 - assert_select "osm>changeset[min_lat='1.0000000']", 1 - assert_select "osm>changeset[max_lat='2.0000000']", 1 + assert_select "osm>changeset[min_lon='0.1000000']", 1 + assert_select "osm>changeset[max_lon='0.2000000']", 1 + assert_select "osm>changeset[min_lat='0.1000000']", 1 + assert_select "osm>changeset[max_lat='0.2000000']", 1 # add (delete) a way to it, which contains a point at (3,3) with_controller(WaysController.new) do @@ -2084,10 +2153,10 @@ module Api # get the bounding box back from the changeset get changeset_show_path(:id => changeset_id) assert_response :success, "Couldn't read back changeset for the third time." - assert_select "osm>changeset[min_lon='1.0000000']", 1 - assert_select "osm>changeset[max_lon='3.0000000']", 1 - assert_select "osm>changeset[min_lat='1.0000000']", 1 - assert_select "osm>changeset[max_lat='3.0000000']", 1 + assert_select "osm>changeset[min_lon='0.1000000']", 1 + assert_select "osm>changeset[max_lon='0.3000000']", 1 + assert_select "osm>changeset[min_lat='0.1000000']", 1 + assert_select "osm>changeset[max_lat='0.3000000']", 1 end ## diff --git a/test/controllers/api/old_nodes_controller_test.rb b/test/controllers/api/old_nodes_controller_test.rb index badc7301b..7855079cc 100644 --- a/test/controllers/api/old_nodes_controller_test.rb +++ b/test/controllers/api/old_nodes_controller_test.rb @@ -40,9 +40,9 @@ module Api # FIXME: Move this test to being an integration test since it spans multiple controllers def test_version private_user = create(:user, :data_public => false) - private_node = create(:node, :with_history, :version => 4, :changeset => create(:changeset, :user => private_user)) + private_node = create(:node, :with_history, :version => 4, :lat => 0, :lon => 0, :changeset => create(:changeset, :user => private_user)) user = create(:user) - node = create(:node, :with_history, :version => 4, :changeset => create(:changeset, :user => user)) + node = create(:node, :with_history, :version => 4, :lat => 0, :lon => 0, :changeset => create(:changeset, :user => user)) create_list(:node_tag, 2, :node => node) # Ensure that the current tags are propagated to the history too propagate_tags(node, node.old_nodes.last) @@ -65,8 +65,8 @@ module Api # randomly move the node about 3.times do # move the node somewhere else - xml_node["lat"] = precision((rand * 180) - 90).to_s - xml_node["lon"] = precision((rand * 360) - 180).to_s + xml_node["lat"] = precision(rand - 0.5).to_s + xml_node["lon"] = precision(rand - 0.5).to_s with_controller(NodesController.new) do put api_node_path(nodeid), :params => xml_doc.to_s, :headers => auth_header assert_response :forbidden, "Should have rejected node update" @@ -113,8 +113,8 @@ module Api # randomly move the node about 3.times do # move the node somewhere else - xml_node["lat"] = precision((rand * 180) - 90).to_s - xml_node["lon"] = precision((rand * 360) - 180).to_s + xml_node["lat"] = precision(rand - 0.5).to_s + xml_node["lon"] = precision(rand - 0.5).to_s with_controller(NodesController.new) do put api_node_path(nodeid), :params => xml_doc.to_s, :headers => auth_header assert_response :success diff --git a/test/controllers/api/relations_controller_test.rb b/test/controllers/api/relations_controller_test.rb index e69dcdfe2..982df1dd7 100644 --- a/test/controllers/api/relations_controller_test.rb +++ b/test/controllers/api/relations_controller_test.rb @@ -641,15 +641,15 @@ module Api # box of all its members into the changeset. def test_tag_modify_bounding_box relation = create(:relation) - node1 = create(:node, :lat => 3, :lon => 3) - node2 = create(:node, :lat => 5, :lon => 5) + node1 = create(:node, :lat => 0.3, :lon => 0.3) + node2 = create(:node, :lat => 0.5, :lon => 0.5) way = create(:way) create(:way_node, :way => way, :node => node1) create(:relation_member, :relation => relation, :member => way) create(:relation_member, :relation => relation, :member => node2) # the relation contains nodes1 and node2 (node1 - # indirectly via the way), so the bbox should be [3,3,5,5]. - check_changeset_modify(BoundingBox.new(3, 3, 5, 5)) do |changeset_id, auth_header| + # indirectly via the way), so the bbox should be [0.3,0.3,0.5,0.5]. + check_changeset_modify(BoundingBox.new(0.3, 0.3, 0.5, 0.5)) do |changeset_id, auth_header| # add a tag to an existing relation relation_xml = xml_for_relation(relation) relation_element = relation_xml.find("//osm/relation").first @@ -879,14 +879,14 @@ module Api # still technically valid. def test_remove_all_members relation = create(:relation) - node1 = create(:node, :lat => 3, :lon => 3) - node2 = create(:node, :lat => 5, :lon => 5) + node1 = create(:node, :lat => 0.3, :lon => 0.3) + node2 = create(:node, :lat => 0.5, :lon => 0.5) way = create(:way) create(:way_node, :way => way, :node => node1) create(:relation_member, :relation => relation, :member => way) create(:relation_member, :relation => relation, :member => node2) - check_changeset_modify(BoundingBox.new(3, 3, 5, 5)) do |changeset_id, auth_header| + check_changeset_modify(BoundingBox.new(0.3, 0.3, 0.5, 0.5)) do |changeset_id, auth_header| relation_xml = xml_for_relation(relation) relation_xml .find("//osm/relation/member") diff --git a/test/factories/node.rb b/test/factories/node.rb index 392d67a84..bfb8b16fe 100644 --- a/test/factories/node.rb +++ b/test/factories/node.rb @@ -16,7 +16,7 @@ FactoryBot.define do trait :with_history do after(:create) do |node, _evaluator| (1..node.version).each do |n| - create(:old_node, :node_id => node.id, :version => n, :changeset => node.changeset) + create(:old_node, :node_id => node.id, :version => n, :latitude => node.latitude, :longitude => node.longitude, :changeset => node.changeset) end # For deleted nodes, make sure the most recent old_node is also deleted. diff --git a/test/models/node_test.rb b/test/models/node_test.rb index 94cb5ec81..9efe9a9c9 100644 --- a/test/models/node_test.rb +++ b/test/models/node_test.rb @@ -98,8 +98,9 @@ class NodeTest < ActiveSupport::TestCase end def test_update - node = create(:node) - create(:old_node, :node_id => node.id, :version => 1) + node = create(:node, :lat => 12.6543, :lon => 65.1234) + create(:old_node, :node_id => node.id, :version => 1, :lat => node.lat, :lon => node.lon) + node_template = Node.find(node.id) assert_not_nil node_template