akkoma/lib/pleroma/object/fetcher.ex

379 lines
12 KiB
Elixir
Raw Normal View History

# Pleroma: A lightweight social networking server
# Copyright © 2017-2021 Pleroma Authors <https://pleroma.social/>
# SPDX-License-Identifier: AGPL-3.0-only
defmodule Pleroma.Object.Fetcher do
2019-05-24 22:24:21 -06:00
alias Pleroma.HTTP
2022-11-15 10:23:47 -07:00
alias Pleroma.Instances
2020-09-15 09:22:08 -06:00
alias Pleroma.Maps
alias Pleroma.Object
alias Pleroma.Object.Containment
alias Pleroma.Repo
alias Pleroma.Signature
alias Pleroma.Web.ActivityPub.InternalFetchActor
alias Pleroma.Web.ActivityPub.ObjectValidator
alias Pleroma.Web.ActivityPub.Transmogrifier
alias Pleroma.Web.Federator
require Logger
require Pleroma.Constants
2024-03-13 14:42:51 -06:00
@moduledoc """
This module deals with correctly fetching Acitivity Pub objects in a safe way.
The core function is `fetch_and_contain_remote_object_from_id/1` which performs
the actual fetch and common safety and authenticity checks. Other `fetch_*`
function use the former and perform some additional tasks
"""
@mix_env Mix.env()
defp touch_changeset(changeset) do
updated_at =
NaiveDateTime.utc_now()
|> NaiveDateTime.truncate(:second)
Ecto.Changeset.put_change(changeset, :updated_at, updated_at)
end
defp maybe_reinject_internal_fields(%{data: %{} = old_data}, new_data) do
has_history? = fn
%{"formerRepresentations" => %{"orderedItems" => list}} when is_list(list) -> true
_ -> false
end
internal_fields = Map.take(old_data, Pleroma.Constants.object_internal_fields())
remote_history_exists? = has_history?.(new_data)
# If the remote history exists, we treat that as the only source of truth.
new_data =
if has_history?.(old_data) and not remote_history_exists? do
Map.put(new_data, "formerRepresentations", old_data["formerRepresentations"])
else
new_data
end
# If the remote does not have history information, we need to manage it ourselves
new_data =
if not remote_history_exists? do
changed? =
Pleroma.Constants.status_updatable_fields()
|> Enum.any?(fn field -> Map.get(old_data, field) != Map.get(new_data, field) end)
%{updated_object: updated_object} =
new_data
|> Object.Updater.maybe_update_history(old_data,
updated: changed?,
use_history_in_new_object?: false
)
updated_object
else
new_data
end
Map.merge(new_data, internal_fields)
end
defp maybe_reinject_internal_fields(_, new_data), do: new_data
@spec reinject_object(struct(), map()) :: {:ok, Object.t()} | {:error, any()}
defp reinject_object(%Object{data: %{"type" => "Question"}} = object, new_data) do
Logger.debug("Reinjecting object #{new_data["id"]}")
with data <- maybe_reinject_internal_fields(object, new_data),
{:ok, data, _} <- ObjectValidator.validate(data, %{}),
changeset <- Object.change(object, %{data: data}),
changeset <- touch_changeset(changeset),
{:ok, object} <- Repo.insert_or_update(changeset),
{:ok, object} <- Object.set_cache(object) do
{:ok, object}
else
e ->
Logger.error("Error while processing object: #{inspect(e)}")
{:error, e}
end
end
defp reinject_object(%Object{} = object, new_data) do
Logger.debug("Reinjecting object #{new_data["id"]}")
with new_data <- Transmogrifier.fix_object(new_data),
data <- maybe_reinject_internal_fields(object, new_data),
changeset <- Object.change(object, %{data: data}),
changeset <- touch_changeset(changeset),
{:ok, object} <- Repo.insert_or_update(changeset),
{:ok, object} <- Object.set_cache(object) do
{:ok, object}
else
e ->
Logger.error("Error while processing object: #{inspect(e)}")
{:error, e}
end
end
2024-03-13 14:42:51 -06:00
@doc "Assumes object already is in our database and refetches from remote to update (e.g. for polls)"
def refetch_object(%Object{data: %{"id" => id}} = object) do
with {:local, false} <- {:local, Object.local?(object)},
{:ok, new_data} <- fetch_and_contain_remote_object_from_id(id),
{:id, true} <- {:id, new_data["id"] == id},
{:ok, object} <- reinject_object(object, new_data) do
{:ok, object}
else
{:local, true} -> {:ok, object}
{:id, false} -> {:error, "Object id changed on refetch"}
e -> {:error, e}
end
end
2024-03-13 14:42:51 -06:00
@doc """
Fetches a new object and puts it through the processing pipeline for inbound objects
Note: will also insert a fake Create activity, since atm we internally
need everything to be traced back to a Create activity.
"""
def fetch_object_from_id(id, options \\ []) do
with %URI{} = uri <- URI.parse(id),
# let's check the URI is even vaguely valid first
{:scheme, true} <- {:scheme, uri.scheme == "http" or uri.scheme == "https"},
# If we have instance restrictions, apply them here to prevent fetching from unwanted instances
{:ok, nil} <- Pleroma.Web.ActivityPub.MRF.SimplePolicy.check_reject(uri),
{:ok, _} <- Pleroma.Web.ActivityPub.MRF.SimplePolicy.check_accept(uri),
{_, nil} <- {:fetch_object, Object.get_cached_by_ap_id(id)},
{_, true} <- {:allowed_depth, Federator.allowed_thread_distance?(options[:depth])},
{_, {:ok, data}} <- {:fetch, fetch_and_contain_remote_object_from_id(id)},
{_, nil} <- {:normalize, Object.normalize(data, fetch: false)},
2019-09-10 22:23:33 -06:00
params <- prepare_activity_params(data),
{_, {:ok, activity}} <-
2019-10-17 22:08:25 -06:00
{:transmogrifier, Transmogrifier.handle_incoming(params, options)},
{_, _data, %Object{} = object} <-
{:object, data, Object.normalize(activity, fetch: false)} do
{:ok, object}
else
{:allowed_depth, false} ->
{:error, "Max thread distance exceeded."}
{:scheme, false} ->
{:error, "URI Scheme Invalid"}
{:transmogrifier, {:error, {:reject, e}}} ->
{:reject, e}
{:transmogrifier, {:reject, e}} ->
{:reject, e}
{:transmogrifier, _} = e ->
{:error, e}
2019-09-10 22:23:33 -06:00
{:object, data, nil} ->
reinject_object(%Object{}, data)
2019-09-10 22:23:33 -06:00
{:normalize, object = %Object{}} ->
{:ok, object}
2019-09-10 22:23:33 -06:00
{:fetch_object, %Object{} = object} ->
{:ok, object}
2019-10-24 10:08:34 -06:00
{:fetch, {:error, error}} ->
{:error, error}
{:reject, reason} ->
{:reject, reason}
2019-10-17 17:37:21 -06:00
e ->
e
end
2019-09-10 22:23:33 -06:00
end
defp prepare_activity_params(data) do
%{
"type" => "Create",
# Should we seriously keep this attributedTo thing?
"actor" => data["actor"] || data["attributedTo"],
"object" => data
}
2020-09-15 09:22:08 -06:00
|> Maps.put_if_present("to", data["to"])
|> Maps.put_if_present("cc", data["cc"])
|> Maps.put_if_present("bto", data["bto"])
|> Maps.put_if_present("bcc", data["bcc"])
end
2024-03-13 14:42:51 -06:00
@doc "Identical to `fetch_object_from_id/2` but just directly returns the object or on error `nil`"
def fetch_object_from_id!(id, options \\ []) do
with {:ok, object} <- fetch_object_from_id(id, options) do
object
else
2019-10-24 10:08:34 -06:00
{:error, %Tesla.Mock.Error{}} ->
nil
{:error, {"Object has been deleted", _id, _code}} ->
nil
2020-07-13 06:23:03 -06:00
{:reject, reason} ->
Logger.debug("Rejected #{id} while fetching: #{inspect(reason)}")
2020-07-13 06:23:03 -06:00
nil
e ->
Logger.error("Error while fetching #{id}: #{inspect(e)}")
nil
end
end
defp make_signature(id, date) do
uri = URI.parse(id)
signature =
InternalFetchActor.get_actor()
|> Signature.sign(%{
"(request-target)": "get #{uri.path}",
host: uri.host,
date: date
})
{"signature", signature}
end
defp sign_fetch(headers, id, date) do
if Pleroma.Config.get([:activitypub, :sign_object_fetches]) do
[make_signature(id, date) | headers]
else
headers
end
end
defp maybe_date_fetch(headers, date) do
if Pleroma.Config.get([:activitypub, :sign_object_fetches]) do
[{"date", date} | headers]
else
headers
end
end
2024-03-13 14:42:51 -06:00
@doc "Fetches arbitrary remote object and performs basic safety and authenticity checks"
def fetch_and_contain_remote_object_from_id(id)
def fetch_and_contain_remote_object_from_id(%{"id" => id}),
do: fetch_and_contain_remote_object_from_id(id)
def fetch_and_contain_remote_object_from_id(id) when is_binary(id) do
2019-12-10 01:08:57 -07:00
Logger.debug("Fetching object #{id} via AP")
with {:scheme, true} <- {:scheme, String.starts_with?(id, "http")},
{_, :ok} <- {:local_fetch, Containment.contain_local_fetch(id)},
{:ok, final_id, body} <- get_object(id),
{:ok, data} <- safe_json_decode(body),
Only allow exact id matches This protects us from falling for obvious spoofs as from the current upload exploit (unfortunately we can’t reasonably do anything about spoofs with exact matches as was possible via emoji and proxy). Such objects being invalid is supported by the spec, sepcifically sections 3.1 and 3.2: https://www.w3.org/TR/activitypub/#obj-id Anonymous objects are not relevant here (they can only exists within parent objects iiuc) and neither is client-to-server or transient objects (as those cannot be fetched in the first place). This leaves us with the requirement for `id` to (a) exist and (b) be a publicly dereferencable URI from the originating server. This alone does not yet demand strict equivalence, but the spec then further explains objects ought to be fetchable _via their ID_. Meaning an object not retrievable via its ID, is invalid. This reading is supported by the fact, e.g. GoToSocial (recently) and Mastodon (for 6+ years) do already implement such strict ID checks, additionally proving this doesn’t cause federation issues in practice. However, apart from canonical IDs there can also be additional display URLs. *omas first redirect those to their canonical location, but *keys and Mastodon directly serve the AP representation without redirects. Mastodon and GTS deal with this in two different ways, but both constitute an effective countermeasure: - Mastodon: Unless it already is a known AP id, two fetches occur. The first fetch just reads the `id` property and then refetches from the id. The last fetch requires the returned id to exactly match the URL the content was fetched from. (This can be optimised by skipping the second fetch if it already matches) https://github.com/mastodon/mastodon/blob/05eda8d19330a9c27c0cf07de19a87edff269057/app/helpers/jsonld_helper.rb#L168 https://github.com/mastodon/mastodon/commit/63f097979990bf5ba9db848b8a253056bad781af - GTS: Only does a single fetch and then checks if _either_ the id _or_ url property (which can be an object) match the original fetch URL. This relies on implementations always including their display URL as "url" if differing from the id. For actors this is true for all investigated implementations, for posts only Mastodon includes an "url", but it is also the only one with a differing display URL. https://github.com/superseriousbusiness/gotosocial/commit/2bafd7daf542d985ee76d9079a30a602cb7be827#diff-943bbb02c8ac74ac5dc5d20807e561dcdfaebdc3b62b10730f643a20ac23c24fR222 Albeit Mastodon’s refetch offers higher compatibility with theoretical implmentations using either multiple different display URL or not denoting any of them as "url" at all, for now we chose to adopt a GTS-like refetch-free approach to avoid additional implementation concerns wrt to whether redirects should be allowed when fetching a canonical AP id and potential for accidentally loosening some checks (e.g. cross-domain refetches) for one of the fetches. This may be reconsidered in the future.
2024-03-15 18:00:19 -06:00
{_, :ok} <- {:strict_id, Containment.contain_id_to_fetch(final_id, data)},
{_, :ok} <- {:containment, Containment.contain_origin(final_id, data)} do
unless Instances.reachable?(final_id) do
Instances.set_reachable(final_id)
2022-11-15 10:23:47 -07:00
end
{:ok, data}
else
Only allow exact id matches This protects us from falling for obvious spoofs as from the current upload exploit (unfortunately we can’t reasonably do anything about spoofs with exact matches as was possible via emoji and proxy). Such objects being invalid is supported by the spec, sepcifically sections 3.1 and 3.2: https://www.w3.org/TR/activitypub/#obj-id Anonymous objects are not relevant here (they can only exists within parent objects iiuc) and neither is client-to-server or transient objects (as those cannot be fetched in the first place). This leaves us with the requirement for `id` to (a) exist and (b) be a publicly dereferencable URI from the originating server. This alone does not yet demand strict equivalence, but the spec then further explains objects ought to be fetchable _via their ID_. Meaning an object not retrievable via its ID, is invalid. This reading is supported by the fact, e.g. GoToSocial (recently) and Mastodon (for 6+ years) do already implement such strict ID checks, additionally proving this doesn’t cause federation issues in practice. However, apart from canonical IDs there can also be additional display URLs. *omas first redirect those to their canonical location, but *keys and Mastodon directly serve the AP representation without redirects. Mastodon and GTS deal with this in two different ways, but both constitute an effective countermeasure: - Mastodon: Unless it already is a known AP id, two fetches occur. The first fetch just reads the `id` property and then refetches from the id. The last fetch requires the returned id to exactly match the URL the content was fetched from. (This can be optimised by skipping the second fetch if it already matches) https://github.com/mastodon/mastodon/blob/05eda8d19330a9c27c0cf07de19a87edff269057/app/helpers/jsonld_helper.rb#L168 https://github.com/mastodon/mastodon/commit/63f097979990bf5ba9db848b8a253056bad781af - GTS: Only does a single fetch and then checks if _either_ the id _or_ url property (which can be an object) match the original fetch URL. This relies on implementations always including their display URL as "url" if differing from the id. For actors this is true for all investigated implementations, for posts only Mastodon includes an "url", but it is also the only one with a differing display URL. https://github.com/superseriousbusiness/gotosocial/commit/2bafd7daf542d985ee76d9079a30a602cb7be827#diff-943bbb02c8ac74ac5dc5d20807e561dcdfaebdc3b62b10730f643a20ac23c24fR222 Albeit Mastodon’s refetch offers higher compatibility with theoretical implmentations using either multiple different display URL or not denoting any of them as "url" at all, for now we chose to adopt a GTS-like refetch-free approach to avoid additional implementation concerns wrt to whether redirects should be allowed when fetching a canonical AP id and potential for accidentally loosening some checks (e.g. cross-domain refetches) for one of the fetches. This may be reconsidered in the future.
2024-03-15 18:00:19 -06:00
{:strict_id, _} ->
{:error, "Object's ActivityPub id/url does not match final fetch URL"}
{:scheme, _} ->
{:error, "Unsupported URI scheme"}
{:local_fetch, _} ->
{:error, "Trying to fetch local resource"}
{:containment, _} ->
{:error, "Object containment failed."}
{:error, e} ->
{:error, e}
e ->
{:error, e}
end
end
def fetch_and_contain_remote_object_from_id(_id),
do: {:error, "id must be a string"}
defp check_crossdomain_redirect(final_host, original_url)
# HOPEFULLY TEMPORARY
# Basically none of our Tesla mocks in tests set the (supposed to
# exist for Tesla proper) url parameter for their responses
# causing almost every fetch in test to fail otherwise
if @mix_env == :test do
defp check_crossdomain_redirect(nil, _) do
{:cross_domain_redirect, false}
end
end
defp check_crossdomain_redirect(final_host, original_url) do
{:cross_domain_redirect, final_host != URI.parse(original_url).host}
end
if @mix_env == :test do
defp get_final_id(nil, initial_url), do: initial_url
defp get_final_id("", initial_url), do: initial_url
end
defp get_final_id(final_url, _intial_url) do
final_url
end
2024-03-13 14:42:51 -06:00
@doc "Do NOT use; only public for use in tests"
def get_object(id) do
2019-08-22 13:39:06 -06:00
date = Pleroma.Signature.signed_date()
headers =
[
# The first is required by spec, the second provided as a fallback for buggy implementations
{"accept", "application/ld+json; profile=\"https://www.w3.org/ns/activitystreams\""},
{"accept", "application/activity+json"}
]
|> maybe_date_fetch(date)
|> sign_fetch(id, date)
with {:ok, %{body: body, status: code, headers: headers, url: final_url}}
when code in 200..299 <-
HTTP.get(id, headers),
remote_host <-
URI.parse(final_url).host,
{:cross_domain_redirect, false} <-
check_crossdomain_redirect(remote_host, id),
{:has_content_type, {_, content_type}} <-
{:has_content_type, List.keyfind(headers, "content-type", 0)},
{:parse_content_type, {:ok, "application", subtype, type_params}} <-
{:parse_content_type, Plug.Conn.Utils.media_type(content_type)} do
final_id = get_final_id(final_url, id)
case {subtype, type_params} do
{"activity+json", _} ->
{:ok, final_id, body}
{"ld+json", %{"profile" => "https://www.w3.org/ns/activitystreams"}} ->
{:ok, final_id, body}
_ ->
{:error, {:content_type, content_type}}
end
else
2019-06-13 04:13:35 -06:00
{:ok, %{status: code}} when code in [404, 410] ->
{:error, {"Object has been deleted", id, code}}
2019-06-13 03:34:03 -06:00
2019-10-24 10:08:34 -06:00
{:error, e} ->
{:error, e}
{:has_content_type, _} ->
{:error, {:content_type, nil}}
{:parse_content_type, e} ->
{:error, {:content_type, e}}
2019-10-17 20:42:25 -06:00
e ->
{:error, e}
end
end
defp safe_json_decode(nil), do: {:ok, nil}
defp safe_json_decode(json), do: Jason.decode(json)
end