Rewrite images to use local proxy (#4035)

* Add markdown rule to add rel=nofollow for all links

* Add markdown image rule to add local image proxy (fixes #1036)

* comments

* rewrite markdown image links working

* add comment

* perform markdown image processing in api/apub receivers

* clippy

* add db table to validate proxied links

* rewrite link fields for avatar, banner etc

* sql fmt

* proxy links received over federation

* add config option

* undo post.url rewriting, move http route definition

* add tests

* proxy images through pictrs

* testing

* cleanup request.rs file

* more cleanup (fixes #2611)

* include url content type when sending post over apub (fixes #2611)

* store post url content type in db

* should be media_type

* get rid of cache_remote_thumbnails setting, instead automatically
take thumbnail from federation data if available.

* fix tests

* add setting disable_external_link_previews

* federate post url as image depending on mime type

* change setting again

* machete

* invert

* support custom emoji

* clippy

* update defaults

* add image proxy test, fix test

* fix test

* clippy

* revert accidental changes

* address review

* clippy

* Markdown link rule-dess (#4356)

* Extracting opengraph_data to its own type.

* A few additions for markdown-link-rule.

---------

Co-authored-by: Nutomic <[email protected]>

* fix setting

* use enum for image proxy setting

* fix test configs

* add config backwards compat

* clippy

* machete

---------

Co-authored-by: Dessalines <[email protected]>

Cargo.toml

@@ -146,7 +146,6 @@ strum_macros = "0.25.3"
 itertools = "0.12.0"
 futures = "0.3.30"
 http = "0.2.11"
-percent-encoding = "2.3.1"
 rosetta-i18n = "0.1.3"
 opentelemetry = { version = "0.19.0", features = ["rt-tokio"] }
 tracing-opentelemetry = { version = "0.19.0" }
@@ -155,6 +154,7 @@ rustls = { version = "0.21.10", features = ["dangerous_configuration"] }
 futures-util = "0.3.30"
 tokio-postgres = "0.7.10"
 tokio-postgres-rustls = "0.10.0"
+urlencoding = "2.1.3"
 enum-map = "2.7"
 moka = { version = "0.12.4", features = ["future"] }
 i-love-jesus = { version = "0.1.0" }

api_tests/src/image.spec.ts

@@ -7,16 +7,23 @@ import {
   PurgePost,
 } from "lemmy-js-client";
 import {
+  alpha,
   alphaImage,
   alphaUrl,
   beta,
   betaUrl,
+  createCommunity,
   createPost,
+  delta,
+  epsilon,
+  gamma,
   getSite,
   registerUser,
   resolveBetaCommunity,
+  resolvePost,
   setupLogins,
   unfollowRemotes,
+  waitForPost,
 } from "./shared";
 const downloadFileSync = require("download-file-sync");
 
@@ -29,9 +36,8 @@ afterAll(() => {
 test("Upload image and delete it", async () => {
   // Upload test image. We use a simple string buffer as pictrs doesnt require an actual image
   // in testing mode.
-  const upload_image = Buffer.from("test");
   const upload_form: UploadImage = {
-    image: upload_image,
+    image: Buffer.from("test"),
   };
   const upload = await alphaImage.uploadImage(upload_form);
   expect(upload.files![0].file).toBeDefined();
@@ -60,9 +66,8 @@ test("Purge user, uploaded image removed", async () => {
   let user = await registerUser(alphaImage, alphaUrl);
 
   // upload test image
-  const upload_image = Buffer.from("test");
   const upload_form: UploadImage = {
-    image: upload_image,
+    image: Buffer.from("test"),
   };
   const upload = await user.uploadImage(upload_form);
   expect(upload.files![0].file).toBeDefined();
@@ -91,9 +96,8 @@ test("Purge post, linked image removed", async () => {
   let user = await registerUser(beta, betaUrl);
 
   // upload test image
-  const upload_image = Buffer.from("test");
   const upload_form: UploadImage = {
-    image: upload_image,
+    image: Buffer.from("test"),
   };
   const upload = await user.uploadImage(upload_form);
   expect(upload.files![0].file).toBeDefined();
@@ -124,3 +128,79 @@ test("Purge post, linked image removed", async () => {
   const content2 = downloadFileSync(upload.url);
   expect(content2).toBe("");
 });
+
+test("Images in remote post are proxied if setting enabled", async () => {
+  let user = await registerUser(beta, betaUrl);
+  let community = await createCommunity(gamma);
+
+  const upload_form: UploadImage = {
+    image: Buffer.from("test"),
+  };
+  const upload = await user.uploadImage(upload_form);
+  let post = await createPost(
+    gamma,
+    community.community_view.community.id,
+    upload.url,
+    "![](http://example.com/image2.png)",
+  );
+  expect(post.post_view.post).toBeDefined();
+
+  // remote image gets proxied after upload
+  expect(
+    post.post_view.post.url?.startsWith(
+      "http://lemmy-gamma:8561/api/v3/image_proxy?url",
+    ),
+  ).toBeTruthy();
+  expect(
+    post.post_view.post.body?.startsWith(
+      "![](http://lemmy-gamma:8561/api/v3/image_proxy?url",
+    ),
+  ).toBeTruthy();
+
+  let epsilonPost = await resolvePost(epsilon, post.post_view.post);
+  expect(epsilonPost.post).toBeDefined();
+
+  // remote image gets proxied after federation
+  expect(
+    epsilonPost.post!.post.url?.startsWith(
+      "http://lemmy-epsilon:8581/api/v3/image_proxy?url",
+    ),
+  ).toBeTruthy();
+  expect(
+    epsilonPost.post!.post.body?.startsWith(
+      "![](http://lemmy-epsilon:8581/api/v3/image_proxy?url",
+    ),
+  ).toBeTruthy();
+});
+
+test("No image proxying if setting is disabled", async () => {
+  let user = await registerUser(beta, betaUrl);
+  let community = await createCommunity(alpha);
+
+  const upload_form: UploadImage = {
+    image: Buffer.from("test"),
+  };
+  const upload = await user.uploadImage(upload_form);
+  let post = await createPost(
+    alpha,
+    community.community_view.community.id,
+    upload.url,
+    "![](http://example.com/image2.png)",
+  );
+  expect(post.post_view.post).toBeDefined();
+
+  // remote image doesnt get proxied after upload
+  expect(
+    post.post_view.post.url?.startsWith("http://127.0.0.1:8551/pictrs/image/"),
+  ).toBeTruthy();
+  expect(post.post_view.post.body).toBe("![](http://example.com/image2.png)");
+
+  let gammaPost = await resolvePost(delta, post.post_view.post);
+  expect(gammaPost.post).toBeDefined();
+
+  // remote image doesnt get proxied after federation
+  expect(
+    gammaPost.post!.post.url?.startsWith("http://127.0.0.1:8551/pictrs/image/"),
+  ).toBeTruthy();
+  expect(gammaPost.post!.post.body).toBe("![](http://example.com/image2.png)");
+});

api_tests/src/post.spec.ts

@@ -39,7 +39,7 @@ import {
   loginUser,
 } from "./shared";
 import { PostView } from "lemmy-js-client/dist/types/PostView";
-import { ResolveObject } from "lemmy-js-client";
+import { EditSite, ResolveObject } from "lemmy-js-client";
 
 let betaCommunity: CommunityView | undefined;
 
@@ -72,6 +72,16 @@ function assertPostFederation(postOne?: PostView, postTwo?: PostView) {
 }
 
 test("Create a post", async () => {
+  // Setup some allowlists and blocklists
+  let editSiteForm: EditSite = {
+    allowed_instances: ["lemmy-beta"],
+  };
+  await delta.editSite(editSiteForm);
+
+  editSiteForm.allowed_instances = [];
+  editSiteForm.blocked_instances = ["lemmy-alpha"];
+  await epsilon.editSite(editSiteForm);
+
   if (!betaCommunity) {
     throw "Missing beta community";
   }
@@ -109,6 +119,12 @@ test("Create a post", async () => {
   await expect(
     resolvePost(epsilon, postRes.post_view.post),
   ).rejects.toStrictEqual(Error("couldnt_find_object"));
+
+  // remove added allow/blocklists
+  editSiteForm.allowed_instances = [];
+  editSiteForm.blocked_instances = [];
+  await delta.editSite(editSiteForm);
+  await epsilon.editSite(editSiteForm);
 });
 
 test("Create a post in a non-existent community", async () => {

api_tests/src/shared.ts

@@ -177,13 +177,6 @@ export async function setupLogins() {
   ];
   await gamma.editSite(editSiteForm);
 
-  editSiteForm.allowed_instances = ["lemmy-beta"];
-  await delta.editSite(editSiteForm);
-
-  editSiteForm.allowed_instances = [];
-  editSiteForm.blocked_instances = ["lemmy-alpha"];
-  await epsilon.editSite(editSiteForm);
-
   // Create the main alpha/beta communities
   // Ignore thrown errors of duplicates
   try {
@@ -203,10 +196,10 @@ export async function createPost(
   api: LemmyHttp,
   community_id: number,
   url: string = "https://example.com/",
+  body = randomString(10),
   // use example.com for consistent title and embed description
   name: string = randomString(5),
 ): Promise<PostResponse> {
-  let body = randomString(10);
   let form: CreatePost = {
     name,
     url,
@@ -528,7 +521,7 @@ export async function likeComment(
 
 export async function createCommunity(
   api: LemmyHttp,
-  name_: string = randomString(5),
+  name_: string = randomString(10),
 ): Promise<CommunityResponse> {
   let description = "a sample description";
   let form: CreateCommunity = {

config/defaults.hjson

@@ -36,22 +36,41 @@
     # Maximum number of active sql connections
     pool_size: 30
   }
-  # Settings related to activitypub federation
   # Pictrs image server configuration.
   pictrs: {
     # Address where pictrs is available (for image hosting)
     url: "http://localhost:8080/"
     # Set a custom pictrs API key. ( Required for deleting images )
     api_key: "string"
-    # By default the thumbnails for external links are stored in pict-rs. This ensures that they
-    # can be reliably retrieved and can be resized using pict-rs APIs. However it also increases
-    # storage usage. In case this is disabled, the Opengraph image is directly returned as
-    # thumbnail.
+    # Backwards compatibility with 0.18.1. False is equivalent to `image_mode: None`, true is
+    # equivalent to `image_mode: StoreLinkPreviews`.
     # 
-    # In some countries it is forbidden to copy preview images from newspaper articles and only
-    # hotlinking is allowed. If that is the case for your instance, make sure that this setting is
-    # disabled.
+    # To be removed in 0.20
     cache_external_link_previews: true
+    # Specifies how to handle remote images, so that users don't have to connect directly to remote servers.
+    image_mode: 
+      # Leave images unchanged, don't generate any local thumbnails for post urls. Instead the the
+      # Opengraph image is directly returned as thumbnail
+      "None"
+
+      # or
+
+      # Generate thumbnails for external post urls and store them persistently in pict-rs. This
+      # ensures that they can be reliably retrieved and can be resized using pict-rs APIs. However
+      # it also increases storage usage.
+      # 
+      # This is the default behaviour, and also matches Lemmy 0.18.
+      "StoreLinkPreviews"
+
+      # or
+
+      # If enabled, all images from remote domains are rewritten to pass through `/api/v3/image_proxy`,
+      # including embedded images in markdown. Images are stored temporarily in pict-rs for caching.
+      # This improves privacy as users don't expose their IP to untrusted servers, and decreases load
+      # on other servers. However it increases bandwidth use for the local server.
+      # 
+      # Requires pict-rs 0.5
+      "ProxyAllImages"
     # Timeout for uploading images to pictrs (in seconds)
     upload_timeout: 30
   }

crates/api/src/local_user/save_settings.rs

@@ -2,7 +2,12 @@ use actix_web::web::{Data, Json};
 use lemmy_api_common::{
   context::LemmyContext,
   person::SaveUserSettings,
-  utils::send_verification_email,
+  utils::{
+    local_site_to_slur_regex,
+    process_markdown_opt,
+    proxy_image_link_opt_api,
+    send_verification_email,
+  },
   SuccessResponse,
 };
 use lemmy_db_schema::{
@@ -12,7 +17,7 @@ use lemmy_db_schema::{
     person::{Person, PersonUpdateForm},
   },
   traits::Crud,
-  utils::{diesel_option_overwrite, diesel_option_overwrite_to_url},
+  utils::diesel_option_overwrite,
 };
 use lemmy_db_views::structs::{LocalUserView, SiteView};
 use lemmy_utils::{
@@ -28,9 +33,11 @@ pub async fn save_user_settings(
 ) -> Result<Json<SuccessResponse>, LemmyError> {
   let site_view = SiteView::read_local(&mut context.pool()).await?;
 
-  let avatar = diesel_option_overwrite_to_url(&data.avatar)?;
-  let banner = diesel_option_overwrite_to_url(&data.banner)?;
-  let bio = diesel_option_overwrite(data.bio.clone());
+  let slur_regex = local_site_to_slur_regex(&site_view.local_site);
+  let bio = diesel_option_overwrite(process_markdown_opt(&data.bio, &slur_regex, &context).await?);
+
+  let avatar = proxy_image_link_opt_api(&data.avatar, &context).await?;
+  let banner = proxy_image_link_opt_api(&data.banner, &context).await?;
   let display_name = diesel_option_overwrite(data.display_name.clone());
   let matrix_user_id = diesel_option_overwrite(data.matrix_user_id.clone());
   let email_deref = data.email.as_deref().map(str::to_lowercase);