diff options
Diffstat (limited to 'lib')
-rw-r--r-- | lib/pleroma/application.ex | 3 | ||||
-rw-r--r-- | lib/pleroma/plugs/uploaded_media.ex | 16 | ||||
-rw-r--r-- | lib/pleroma/web/media_proxy/invalidation.ex | 29 | ||||
-rw-r--r-- | lib/pleroma/web/media_proxy/invalidations/http.ex | 8 | ||||
-rw-r--r-- | lib/pleroma/web/media_proxy/invalidations/script.ex | 36 | ||||
-rw-r--r-- | lib/pleroma/web/media_proxy/media_proxy.ex | 35 | ||||
-rw-r--r-- | lib/pleroma/web/media_proxy/media_proxy_controller.ex | 3 | ||||
-rw-r--r-- | lib/pleroma/workers/attachments_cleanup_worker.ex | 129 |
8 files changed, 163 insertions, 96 deletions
diff --git a/lib/pleroma/application.ex b/lib/pleroma/application.ex index 9d3d92b38..adebebc7a 100644 --- a/lib/pleroma/application.ex +++ b/lib/pleroma/application.ex @@ -148,7 +148,8 @@ defmodule Pleroma.Application do build_cachex("idempotency", expiration: idempotency_expiration(), limit: 2500), build_cachex("web_resp", limit: 2500), build_cachex("emoji_packs", expiration: emoji_packs_expiration(), limit: 10), - build_cachex("failed_proxy_url", limit: 2500) + build_cachex("failed_proxy_url", limit: 2500), + build_cachex("deleted_urls", default_ttl: :timer.hours(24 * 30), limit: 5_000) ] end diff --git a/lib/pleroma/plugs/uploaded_media.ex b/lib/pleroma/plugs/uploaded_media.ex index 94147e0c4..2f3fde002 100644 --- a/lib/pleroma/plugs/uploaded_media.ex +++ b/lib/pleroma/plugs/uploaded_media.ex @@ -10,6 +10,8 @@ defmodule Pleroma.Plugs.UploadedMedia do import Pleroma.Web.Gettext require Logger + alias Pleroma.Web.MediaProxy + @behaviour Plug # no slashes @path "media" @@ -35,8 +37,7 @@ defmodule Pleroma.Plugs.UploadedMedia do %{query_params: %{"name" => name}} = conn -> name = String.replace(name, "\"", "\\\"") - conn - |> put_resp_header("content-disposition", "filename=\"#{name}\"") + put_resp_header(conn, "content-disposition", "filename=\"#{name}\"") conn -> conn @@ -47,7 +48,8 @@ defmodule Pleroma.Plugs.UploadedMedia do with uploader <- Keyword.fetch!(config, :uploader), proxy_remote = Keyword.get(config, :proxy_remote, false), - {:ok, get_method} <- uploader.get_file(file) do + {:ok, get_method} <- uploader.get_file(file), + false <- media_is_deleted(conn, get_method) do get_media(conn, get_method, proxy_remote, opts) else _ -> @@ -59,6 +61,14 @@ defmodule Pleroma.Plugs.UploadedMedia do def call(conn, _opts), do: conn + defp media_is_deleted(%{request_path: path} = _conn, {:static_dir, _}) do + MediaProxy.in_deleted_urls(Pleroma.Web.base_url() <> path) + end + + defp media_is_deleted(_, {:url, url}), do: MediaProxy.in_deleted_urls(url) + + defp media_is_deleted(_, _), do: false + defp get_media(conn, {:static_dir, directory}, _, opts) do static_opts = Map.get(opts, :static_plug_opts) diff --git a/lib/pleroma/web/media_proxy/invalidation.ex b/lib/pleroma/web/media_proxy/invalidation.ex index c037ff13e..83ff8589c 100644 --- a/lib/pleroma/web/media_proxy/invalidation.ex +++ b/lib/pleroma/web/media_proxy/invalidation.ex @@ -5,22 +5,33 @@ defmodule Pleroma.Web.MediaProxy.Invalidation do @moduledoc false - @callback purge(list(String.t()), map()) :: {:ok, String.t()} | {:error, String.t()} + @callback purge(list(String.t()), Keyword.t()) :: {:ok, list(String.t())} | {:error, String.t()} alias Pleroma.Config + alias Pleroma.Web.MediaProxy - @spec purge(list(String.t())) :: {:ok, String.t()} | {:error, String.t()} + @spec enabled?() :: boolean() + def enabled?, do: Config.get([:media_proxy, :invalidation, :enabled]) + + @spec purge(list(String.t()) | String.t()) :: {:ok, list(String.t())} | {:error, String.t()} def purge(urls) do - [:media_proxy, :invalidation, :enabled] - |> Config.get() - |> do_purge(urls) + prepared_urls = prepare_urls(urls) + + if enabled?() do + do_purge(prepared_urls) + else + {:ok, prepared_urls} + end end - defp do_purge(true, urls) do + defp do_purge(urls) do provider = Config.get([:media_proxy, :invalidation, :provider]) - options = Config.get(provider) - provider.purge(urls, options) + provider.purge(urls, Config.get(provider)) end - defp do_purge(_, _), do: :ok + def prepare_urls(urls) do + urls + |> List.wrap() + |> Enum.map(&MediaProxy.url(&1)) + end end diff --git a/lib/pleroma/web/media_proxy/invalidations/http.ex b/lib/pleroma/web/media_proxy/invalidations/http.ex index 07248df6e..3694b56e8 100644 --- a/lib/pleroma/web/media_proxy/invalidations/http.ex +++ b/lib/pleroma/web/media_proxy/invalidations/http.ex @@ -10,9 +10,9 @@ defmodule Pleroma.Web.MediaProxy.Invalidation.Http do @impl Pleroma.Web.MediaProxy.Invalidation def purge(urls, opts) do - method = Map.get(opts, :method, :purge) - headers = Map.get(opts, :headers, []) - options = Map.get(opts, :options, []) + method = Keyword.get(opts, :method, :purge) + headers = Keyword.get(opts, :headers, []) + options = Keyword.get(opts, :options, []) Logger.debug("Running cache purge: #{inspect(urls)}") @@ -22,7 +22,7 @@ defmodule Pleroma.Web.MediaProxy.Invalidation.Http do end end) - {:ok, "success"} + {:ok, urls} end defp do_purge(method, url, headers, options) do diff --git a/lib/pleroma/web/media_proxy/invalidations/script.ex b/lib/pleroma/web/media_proxy/invalidations/script.ex index 6be782132..d41d647bb 100644 --- a/lib/pleroma/web/media_proxy/invalidations/script.ex +++ b/lib/pleroma/web/media_proxy/invalidations/script.ex @@ -10,32 +10,34 @@ defmodule Pleroma.Web.MediaProxy.Invalidation.Script do require Logger @impl Pleroma.Web.MediaProxy.Invalidation - def purge(urls, %{script_path: script_path} = _options) do + def purge(urls, opts) do args = urls |> List.wrap() |> Enum.uniq() |> Enum.join(" ") - path = Path.expand(script_path) - - Logger.debug("Running cache purge: #{inspect(urls)}, #{path}") - - case do_purge(path, [args]) do - {result, exit_status} when exit_status > 0 -> - Logger.error("Error while cache purge: #{inspect(result)}") - {:error, inspect(result)} - - _ -> - {:ok, "success"} - end + opts + |> Keyword.get(:script_path, nil) + |> do_purge([args]) + |> handle_result(urls) end - def purge(_, _), do: {:error, "not found script path"} - - defp do_purge(path, args) do + defp do_purge(script_path, args) when is_binary(script_path) do + path = Path.expand(script_path) + Logger.debug("Running cache purge: #{inspect(args)}, #{inspect(path)}") System.cmd(path, args) rescue - error -> {inspect(error), 1} + error -> error + end + + defp do_purge(_, _), do: {:error, "not found script path"} + + defp handle_result({_result, 0}, urls), do: {:ok, urls} + defp handle_result({:error, error}, urls), do: handle_result(error, urls) + + defp handle_result(error, _) do + Logger.error("Error while cache purge: #{inspect(error)}") + {:error, inspect(error)} end end diff --git a/lib/pleroma/web/media_proxy/media_proxy.ex b/lib/pleroma/web/media_proxy/media_proxy.ex index b2b524524..59ca217ab 100644 --- a/lib/pleroma/web/media_proxy/media_proxy.ex +++ b/lib/pleroma/web/media_proxy/media_proxy.ex @@ -6,20 +6,53 @@ defmodule Pleroma.Web.MediaProxy do alias Pleroma.Config alias Pleroma.Upload alias Pleroma.Web + alias Pleroma.Web.MediaProxy.Invalidation @base64_opts [padding: false] + @spec in_deleted_urls(String.t()) :: boolean() + def in_deleted_urls(url), do: elem(Cachex.exists?(:deleted_urls_cache, url(url)), 1) + + def remove_from_deleted_urls(urls) when is_list(urls) do + Cachex.execute!(:deleted_urls_cache, fn cache -> + Enum.each(Invalidation.prepare_urls(urls), &Cachex.del(cache, &1)) + end) + end + + def remove_from_deleted_urls(url) when is_binary(url) do + Cachex.del(:deleted_urls_cache, url(url)) + end + + def put_in_deleted_urls(urls) when is_list(urls) do + Cachex.execute!(:deleted_urls_cache, fn cache -> + Enum.each(Invalidation.prepare_urls(urls), &Cachex.put(cache, &1, true)) + end) + end + + def put_in_deleted_urls(url) when is_binary(url) do + Cachex.put(:deleted_urls_cache, url(url), true) + end + def url(url) when is_nil(url) or url == "", do: nil def url("/" <> _ = url), do: url def url(url) do - if disabled?() or local?(url) or whitelisted?(url) do + if disabled?() or not is_url_proxiable?(url) do url else encode_url(url) end end + @spec is_url_proxiable?(String.t()) :: boolean() + def is_url_proxiable?(url) do + if local?(url) or whitelisted?(url) do + false + else + true + end + end + defp disabled?, do: !Config.get([:media_proxy, :enabled], false) defp local?(url), do: String.starts_with?(url, Pleroma.Web.base_url()) diff --git a/lib/pleroma/web/media_proxy/media_proxy_controller.ex b/lib/pleroma/web/media_proxy/media_proxy_controller.ex index 4657a4383..ff0158d83 100644 --- a/lib/pleroma/web/media_proxy/media_proxy_controller.ex +++ b/lib/pleroma/web/media_proxy/media_proxy_controller.ex @@ -14,10 +14,11 @@ defmodule Pleroma.Web.MediaProxy.MediaProxyController do with config <- Pleroma.Config.get([:media_proxy], []), true <- Keyword.get(config, :enabled, false), {:ok, url} <- MediaProxy.decode_url(sig64, url64), + {_, false} <- {:in_deleted_urls, MediaProxy.in_deleted_urls(url)}, :ok <- filename_matches(params, conn.request_path, url) do ReverseProxy.call(conn, url, Keyword.get(config, :proxy_opts, @default_proxy_opts)) else - false -> + error when error in [false, {:in_deleted_urls, true}] -> send_resp(conn, 404, Plug.Conn.Status.reason_phrase(404)) {:error, :invalid_signature} -> diff --git a/lib/pleroma/workers/attachments_cleanup_worker.ex b/lib/pleroma/workers/attachments_cleanup_worker.ex index 49352db2a..4ad19c0fc 100644 --- a/lib/pleroma/workers/attachments_cleanup_worker.ex +++ b/lib/pleroma/workers/attachments_cleanup_worker.ex @@ -23,8 +23,25 @@ defmodule Pleroma.Workers.AttachmentsCleanupWorker do Enum.map(attachment["url"], & &1["href"]) end) - names = Enum.map(attachments, & &1["name"]) + # find all objects for copies of the attachments, name and actor doesn't matter here + hrefs + |> fetch_objects + |> prepare_objects(actor, Enum.map(attachments, & &1["name"])) + |> Enum.reduce({[], []}, fn {href, %{id: id, count: count}}, {ids, hrefs} -> + with 1 <- count do + {ids ++ [id], hrefs ++ [href]} + else + _ -> {ids ++ [id], hrefs} + end + end) + |> do_clean + + {:ok, :success} + end + def perform(%{"op" => "cleanup_attachments", "object" => _object}, _job), do: {:ok, :skip} + + defp do_clean({object_ids, attachment_urls}) do uploader = Pleroma.Config.get([Pleroma.Upload, :uploader]) prefix = @@ -39,68 +56,60 @@ defmodule Pleroma.Workers.AttachmentsCleanupWorker do "/" ) - # find all objects for copies of the attachments, name and actor doesn't matter here - object_ids_and_hrefs = - from(o in Object, - where: - fragment( - "to_jsonb(array(select jsonb_array_elements((?)#>'{url}') ->> 'href' where jsonb_typeof((?)#>'{url}') = 'array'))::jsonb \\?| (?)", - o.data, - o.data, - ^hrefs - ) - ) - # The query above can be time consumptive on large instances until we - # refactor how uploads are stored - |> Repo.all(timeout: :infinity) - # we should delete 1 object for any given attachment, but don't delete - # files if there are more than 1 object for it - |> Enum.reduce(%{}, fn %{ - id: id, - data: %{ - "url" => [%{"href" => href}], - "actor" => obj_actor, - "name" => name - } - }, - acc -> - Map.update(acc, href, %{id: id, count: 1}, fn val -> - case obj_actor == actor and name in names do - true -> - # set id of the actor's object that will be deleted - %{val | id: id, count: val.count + 1} - - false -> - # another actor's object, just increase count to not delete file - %{val | count: val.count + 1} - end - end) - end) - |> Enum.map(fn {href, %{id: id, count: count}} -> - # only delete files that have single instance - with 1 <- count do - href - |> String.trim_leading("#{base_url}/#{prefix}") - |> uploader.delete_file() - - {id, href} - else - _ -> {id, nil} - end - end) - - object_ids = Enum.map(object_ids_and_hrefs, fn {id, _} -> id end) + Enum.each(attachment_urls, fn href -> + href + |> String.trim_leading("#{base_url}/#{prefix}") + |> uploader.delete_file() + end) - from(o in Object, where: o.id in ^object_ids) - |> Repo.delete_all() + delete_objects(object_ids) + end - object_ids_and_hrefs - |> Enum.filter(fn {_, href} -> not is_nil(href) end) - |> Enum.map(&elem(&1, 1)) - |> Pleroma.Web.MediaProxy.Invalidation.purge() + defp delete_objects([_ | _] = object_ids) do + Repo.delete_all(from(o in Object, where: o.id in ^object_ids)) + end - {:ok, :success} + defp delete_objects(_), do: :ok + + # we should delete 1 object for any given attachment, but don't delete + # files if there are more than 1 object for it + def prepare_objects(objects, actor, names) do + objects + |> Enum.reduce(%{}, fn %{ + id: id, + data: %{ + "url" => [%{"href" => href}], + "actor" => obj_actor, + "name" => name + } + }, + acc -> + Map.update(acc, href, %{id: id, count: 1}, fn val -> + case obj_actor == actor and name in names do + true -> + # set id of the actor's object that will be deleted + %{val | id: id, count: val.count + 1} + + false -> + # another actor's object, just increase count to not delete file + %{val | count: val.count + 1} + end + end) + end) end - def perform(%{"op" => "cleanup_attachments", "object" => _object}, _job), do: {:ok, :skip} + def fetch_objects(hrefs) do + from(o in Object, + where: + fragment( + "to_jsonb(array(select jsonb_array_elements((?)#>'{url}') ->> 'href' where jsonb_typeof((?)#>'{url}') = 'array'))::jsonb \\?| (?)", + o.data, + o.data, + ^hrefs + ) + ) + # The query above can be time consumptive on large instances until we + # refactor how uploads are stored + |> Repo.all(timeout: :infinity) + end end |