media.rs

use std::{io::Cursor, net::IpAddr, sync::Arc, time::Duration};

use image::io::Reader as ImgReader;
use reqwest::Url;
use ruma::api::client::{
	error::ErrorKind,
	media::{
		create_content, get_content, get_content_as_filename, get_content_thumbnail, get_media_config,
		get_media_preview,
	},
};
use tracing::{debug, error, info, warn};
use webpage::HTML;

use crate::{
	service::media::{FileMeta, UrlPreviewData},
	services, utils, Error, Result, Ruma,
};

/// generated MXC ID (`media-id`) length
const MXC_LENGTH: usize = 32;

/// # `GET /_matrix/media/v3/config`
///
/// Returns max upload size.
pub async fn get_media_config_route(
	_body: Ruma<get_media_config::v3::Request>,
) -> Result<get_media_config::v3::Response> {
	Ok(get_media_config::v3::Response {
		upload_size: services().globals.max_request_size().into(),
	})
}

/// # `GET /_matrix/media/v3/preview_url`
///
/// Returns URL preview.
pub async fn get_media_preview_route(
	body: Ruma<get_media_preview::v3::Request>,
) -> Result<get_media_preview::v3::Response> {
	let url = &body.url;
	if !url_preview_allowed(url) {
		return Err(Error::BadRequest(ErrorKind::Forbidden, "URL is not allowed to be previewed"));
	}

	if let Ok(preview) = get_url_preview(url).await {
		let res = serde_json::value::to_raw_value(&preview).map_err(|e| {
			error!("Failed to convert UrlPreviewData into a serde json value: {}", e);
			Error::BadRequest(ErrorKind::Unknown, "Unknown error occurred parsing URL preview")
		})?;

		return Ok(get_media_preview::v3::Response::from_raw_value(res));
	}

	Err(Error::BadRequest(
		ErrorKind::LimitExceeded {
			retry_after_ms: Some(Duration::from_secs(5)),
		},
		"Retry later",
	))
}

/// # `POST /_matrix/media/v3/upload`
///
/// Permanently save media in the server.
///
/// - Some metadata will be saved in the database
/// - Media will be saved in the media/ directory
pub async fn create_content_route(body: Ruma<create_content::v3::Request>) -> Result<create_content::v3::Response> {
	let mxc = format!(
		"mxc://{}/{}",
		services().globals.server_name(),
		utils::random_string(MXC_LENGTH)
	);

	services()
		.media
		.create(
			mxc.clone(),
			body.filename.as_ref().map(|filename| "inline; filename=".to_owned() + filename).as_deref(),
			body.content_type.as_deref(),
			&body.file,
		)
		.await?;

	let content_uri = mxc.into();

	Ok(create_content::v3::Response {
		content_uri,
		blurhash: None,
	})
}

/// helper method to fetch remote media from other servers over federation
pub async fn get_remote_content(
	mxc: &str, server_name: &ruma::ServerName, media_id: String, allow_redirect: bool, timeout_ms: Duration,
) -> Result<get_content::v3::Response, Error> {
	// we'll lie to the client and say the blocked server's media was not found and
	// log. the client has no way of telling anyways so this is a security bonus.
	if services().globals.prevent_media_downloads_from().contains(&server_name.to_owned()) {
		info!(
			"Received request for remote media `{}` but server is in our media server blocklist. Returning 404.",
			mxc
		);
		return Err(Error::BadRequest(ErrorKind::NotFound, "Media not found."));
	}

	let content_response = services()
		.sending
		.send_federation_request(
			server_name,
			get_content::v3::Request {
				allow_remote: true,
				server_name: server_name.to_owned(),
				media_id,
				timeout_ms,
				allow_redirect,
			},
		)
		.await?;

	services()
		.media
		.create(
			mxc.to_owned(),
			content_response.content_disposition.as_deref(),
			content_response.content_type.as_deref(),
			&content_response.file,
		)
		.await?;

	Ok(content_response)
}

/// # `GET /_matrix/media/v3/download/{serverName}/{mediaId}`
///
/// Load media from our server or over federation.
///
/// - Only allows federation if `allow_remote` is true
/// - Only redirects if `allow_redirect` is true
/// - Uses client-provided `timeout_ms` if available, else defaults to 20
///   seconds
pub async fn get_content_route(body: Ruma<get_content::v3::Request>) -> Result<get_content::v3::Response> {
	let mxc = format!("mxc://{}/{}", body.server_name, body.media_id);

	if let Some(FileMeta {
		content_disposition,
		content_type,
		file,
	}) = services().media.get(mxc.clone()).await?
	{
		Ok(get_content::v3::Response {
			file,
			content_type,
			content_disposition,
			cross_origin_resource_policy: Some("cross-origin".to_owned()),
		})
	} else if &*body.server_name != services().globals.server_name() && body.allow_remote {
		let remote_content_response = get_remote_content(
			&mxc,
			&body.server_name,
			body.media_id.clone(),
			body.allow_redirect,
			body.timeout_ms,
		)
		.await?;
		Ok(remote_content_response)
	} else {
		Err(Error::BadRequest(ErrorKind::NotFound, "Media not found."))
	}
}

/// # `GET /_matrix/media/v3/download/{serverName}/{mediaId}/{fileName}`
///
/// Load media from our server or over federation, permitting desired filename.
///
/// - Only allows federation if `allow_remote` is true
/// - Only redirects if `allow_redirect` is true
/// - Uses client-provided `timeout_ms` if available, else defaults to 20
///   seconds
pub async fn get_content_as_filename_route(
	body: Ruma<get_content_as_filename::v3::Request>,
) -> Result<get_content_as_filename::v3::Response> {
	let mxc = format!("mxc://{}/{}", body.server_name, body.media_id);

	if let Some(FileMeta {
		content_type,
		file,
		..
	}) = services().media.get(mxc.clone()).await?
	{
		Ok(get_content_as_filename::v3::Response {
			file,
			content_type,
			content_disposition: Some(format!("inline; filename={}", body.filename)),
			cross_origin_resource_policy: Some("cross-origin".to_owned()),
		})
	} else if &*body.server_name != services().globals.server_name() && body.allow_remote {
		let remote_content_response = get_remote_content(
			&mxc,
			&body.server_name,
			body.media_id.clone(),
			body.allow_redirect,
			body.timeout_ms,
		)
		.await?;

		Ok(get_content_as_filename::v3::Response {
			content_disposition: Some(format!("inline: filename={}", body.filename)),
			content_type: remote_content_response.content_type,
			file: remote_content_response.file,
			cross_origin_resource_policy: Some("cross-origin".to_owned()),
		})
	} else {
		Err(Error::BadRequest(ErrorKind::NotFound, "Media not found."))
	}
}

/// # `GET /_matrix/media/v3/thumbnail/{serverName}/{mediaId}`
///
/// Load media thumbnail from our server or over federation.
///
/// - Only allows federation if `allow_remote` is true
/// - Only redirects if `allow_redirect` is true
/// - Uses client-provided `timeout_ms` if available, else defaults to 20
///   seconds
pub async fn get_content_thumbnail_route(
	body: Ruma<get_content_thumbnail::v3::Request>,
) -> Result<get_content_thumbnail::v3::Response> {
	let mxc = format!("mxc://{}/{}", body.server_name, body.media_id);

	if let Some(FileMeta {
		content_type,
		file,
		..
	}) = services()
		.media
		.get_thumbnail(
			mxc.clone(),
			body.width.try_into().map_err(|_| Error::BadRequest(ErrorKind::InvalidParam, "Width is invalid."))?,
			body.height.try_into().map_err(|_| Error::BadRequest(ErrorKind::InvalidParam, "Height is invalid."))?,
		)
		.await?
	{
		Ok(get_content_thumbnail::v3::Response {
			file,
			content_type,
			cross_origin_resource_policy: Some("cross-origin".to_owned()),
		})
	} else if &*body.server_name != services().globals.server_name() && body.allow_remote {
		// we'll lie to the client and say the blocked server's media was not found and
		// log. the client has no way of telling anyways so this is a security bonus.
		if services().globals.prevent_media_downloads_from().contains(&body.server_name.clone()) {
			info!(
				"Received request for remote media `{}` but server is in our media server blocklist. Returning 404.",
				mxc
			);
			return Err(Error::BadRequest(ErrorKind::NotFound, "Media not found."));
		}

		let get_thumbnail_response = services()
			.sending
			.send_federation_request(
				&body.server_name,
				get_content_thumbnail::v3::Request {
					allow_remote: body.allow_remote,
					height: body.height,
					width: body.width,
					method: body.method.clone(),
					server_name: body.server_name.clone(),
					media_id: body.media_id.clone(),
					timeout_ms: body.timeout_ms,
					allow_redirect: body.allow_redirect,
				},
			)
			.await?;

		services()
			.media
			.upload_thumbnail(
				mxc,
				None,
				get_thumbnail_response.content_type.as_deref(),
				body.width.try_into().expect("all UInts are valid u32s"),
				body.height.try_into().expect("all UInts are valid u32s"),
				&get_thumbnail_response.file,
			)
			.await?;

		Ok(get_thumbnail_response)
	} else {
		Err(Error::BadRequest(ErrorKind::NotFound, "Media not found."))
	}
}

async fn download_image(client: &reqwest::Client, url: &str) -> Result<UrlPreviewData> {
	let image = client.get(url).send().await?.bytes().await?;
	let mxc = format!(
		"mxc://{}/{}",
		services().globals.server_name(),
		utils::random_string(MXC_LENGTH)
	);

	services().media.create(mxc.clone(), None, None, &image).await?;

	let (width, height) = match ImgReader::new(Cursor::new(&image)).with_guessed_format() {
		Err(_) => (None, None),
		Ok(reader) => match reader.into_dimensions() {
			Err(_) => (None, None),
			Ok((width, height)) => (Some(width), Some(height)),
		},
	};

	Ok(UrlPreviewData {
		image: Some(mxc),
		image_size: Some(image.len()),
		image_width: width,
		image_height: height,
		..Default::default()
	})
}

async fn download_html(client: &reqwest::Client, url: &str) -> Result<UrlPreviewData> {
	let mut response = client.get(url).send().await?;

	let mut bytes: Vec<u8> = Vec::new();
	while let Some(chunk) = response.chunk().await? {
		bytes.extend_from_slice(&chunk);
		if bytes.len() > services().globals.url_preview_max_spider_size() {
			debug!(
				"Response body from URL {} exceeds url_preview_max_spider_size ({}), not processing the rest of the \
				 response body and assuming our necessary data is in this range.",
				url,
				services().globals.url_preview_max_spider_size()
			);
			break;
		}
	}
	let body = String::from_utf8_lossy(&bytes);
	let html = match HTML::from_string(body.to_string(), Some(url.to_owned())) {
		Ok(html) => html,
		Err(_) => return Err(Error::BadRequest(ErrorKind::Unknown, "Failed to parse HTML")),
	};

	let mut data = match html.opengraph.images.first() {
		None => UrlPreviewData::default(),
		Some(obj) => download_image(client, &obj.url).await?,
	};

	let props = html.opengraph.properties;

	/* use OpenGraph title/description, but fall back to HTML if not available */
	data.title = props.get("title").cloned().or(html.title);
	data.description = props.get("description").cloned().or(html.description);

	Ok(data)
}

fn url_request_allowed(addr: &IpAddr) -> bool {
	// TODO: make this check ip_range_denylist

	// could be implemented with reqwest when it supports IP filtering:
	// https://github.com/seanmonstar/reqwest/issues/1515

	// These checks have been taken from the Rust core/net/ipaddr.rs crate,
	// IpAddr::V4.is_global() and IpAddr::V6.is_global(), as .is_global is not
	// yet stabilized. TODO: Once this is stable, this match can be simplified.
	match addr {
		IpAddr::V4(ip4) => {
			!(ip4.octets()[0] == 0 // "This network"
                || ip4.is_private()
                || (ip4.octets()[0] == 100 && (ip4.octets()[1] & 0b1100_0000 == 0b0100_0000)) // is_shared()
                || ip4.is_loopback()
                || ip4.is_link_local()
                // addresses reserved for future protocols (`192.0.0.0/24`)
                || (ip4.octets()[0] == 192 && ip4.octets()[1] == 0 && ip4.octets()[2] == 0)
                || ip4.is_documentation()
                || (ip4.octets()[0] == 198 && (ip4.octets()[1] & 0xfe) == 18) // is_benchmarking()
                || (ip4.octets()[0] & 240 == 240 && !ip4.is_broadcast()) // is_reserved()
                || ip4.is_broadcast())
		},
		IpAddr::V6(ip6) => {
			!(ip6.is_unspecified()
                || ip6.is_loopback()
                // IPv4-mapped Address (`::ffff:0:0/96`)
                || matches!(ip6.segments(), [0, 0, 0, 0, 0, 0xffff, _, _])
                // IPv4-IPv6 Translat. (`64:ff9b:1::/48`)
                || matches!(ip6.segments(), [0x64, 0xff9b, 1, _, _, _, _, _])
                // Discard-Only Address Block (`100::/64`)
                || matches!(ip6.segments(), [0x100, 0, 0, 0, _, _, _, _])
                // IETF Protocol Assignments (`2001::/23`)
                || (matches!(ip6.segments(), [0x2001, b, _, _, _, _, _, _] if b < 0x200)
                    && !(
                        // Port Control Protocol Anycast (`2001:1::1`)
                        u128::from_be_bytes(ip6.octets()) == 0x2001_0001_0000_0000_0000_0000_0000_0001
                        // Traversal Using Relays around NAT Anycast (`2001:1::2`)
                        || u128::from_be_bytes(ip6.octets()) == 0x2001_0001_0000_0000_0000_0000_0000_0002
                        // AMT (`2001:3::/32`)
                        || matches!(ip6.segments(), [0x2001, 3, _, _, _, _, _, _])
                        // AS112-v6 (`2001:4:112::/48`)
                        || matches!(ip6.segments(), [0x2001, 4, 0x112, _, _, _, _, _])
                        // ORCHIDv2 (`2001:20::/28`)
                        || matches!(ip6.segments(), [0x2001, b, _, _, _, _, _, _] if (0x20..=0x2F).contains(&b))
                    ))
                || ((ip6.segments()[0] == 0x2001) && (ip6.segments()[1] == 0xdb8)) // is_documentation()
                || ((ip6.segments()[0] & 0xfe00) == 0xfc00) // is_unique_local()
                || ((ip6.segments()[0] & 0xffc0) == 0xfe80)) // is_unicast_link_local
		},
	}
}

async fn request_url_preview(url: &str) -> Result<UrlPreviewData> {
	let client = services().globals.url_preview_client();
	let response = client.head(url).send().await?;

	if !response.remote_addr().map_or(false, |a| url_request_allowed(&a.ip())) {
		return Err(Error::BadRequest(
			ErrorKind::Forbidden,
			"Requesting from this address is forbidden",
		));
	}

	let content_type = match response.headers().get(reqwest::header::CONTENT_TYPE).and_then(|x| x.to_str().ok()) {
		Some(ct) => ct,
		None => return Err(Error::BadRequest(ErrorKind::Unknown, "Unknown Content-Type")),
	};
	let data = match content_type {
		html if html.starts_with("text/html") => download_html(&client, url).await?,
		img if img.starts_with("image/") => download_image(&client, url).await?,
		_ => return Err(Error::BadRequest(ErrorKind::Unknown, "Unsupported Content-Type")),
	};

	services().media.set_url_preview(url, &data).await?;

	Ok(data)
}

async fn get_url_preview(url: &str) -> Result<UrlPreviewData> {
	if let Some(preview) = services().media.get_url_preview(url).await {
		return Ok(preview);
	}

	// ensure that only one request is made per URL
	let mutex_request = Arc::clone(services().media.url_preview_mutex.write().await.entry(url.to_owned()).or_default());
	let _request_lock = mutex_request.lock().await;

	match services().media.get_url_preview(url).await {
		Some(preview) => Ok(preview),
		None => request_url_preview(url).await,
	}
}

fn url_preview_allowed(url_str: &str) -> bool {
	let url: Url = match Url::parse(url_str) {
		Ok(u) => u,
		Err(e) => {
			warn!("Failed to parse URL from a str: {}", e);
			return false;
		},
	};

	if ["http", "https"].iter().all(|&scheme| scheme != url.scheme().to_lowercase()) {
		debug!("Ignoring non-HTTP/HTTPS URL to preview: {}", url);
		return false;
	}

	let host = match url.host_str() {
		None => {
			debug!("Ignoring URL preview for a URL that does not have a host (?): {}", url);
			return false;
		},
		Some(h) => h.to_owned(),
	};

	let allowlist_domain_contains = services().globals.url_preview_domain_contains_allowlist();
	let allowlist_domain_explicit = services().globals.url_preview_domain_explicit_allowlist();
	let allowlist_url_contains = services().globals.url_preview_url_contains_allowlist();

	if allowlist_domain_contains.contains(&"*".to_owned())
		|| allowlist_domain_explicit.contains(&"*".to_owned())
		|| allowlist_url_contains.contains(&"*".to_owned())
	{
		debug!("Config key contains * which is allowing all URL previews. Allowing URL {}", url);
		return true;
	}

	if !host.is_empty() {
		if allowlist_domain_explicit.contains(&host) {
			debug!("Host {} is allowed by url_preview_domain_explicit_allowlist (check 1/3)", &host);
			return true;
		}

		if allowlist_domain_contains.iter().any(|domain_s| domain_s.contains(&host.clone())) {
			debug!("Host {} is allowed by url_preview_domain_contains_allowlist (check 2/3)", &host);
			return true;
		}

		if allowlist_url_contains.iter().any(|url_s| url.to_string().contains(&url_s.to_string())) {
			debug!("URL {} is allowed by url_preview_url_contains_allowlist (check 3/3)", &host);
			return true;
		}

		// check root domain if available and if user has root domain checks
		if services().globals.url_preview_check_root_domain() {
			debug!("Checking root domain");
			match host.split_once('.') {
				None => return false,
				Some((_, root_domain)) => {
					if allowlist_domain_explicit.contains(&root_domain.to_owned()) {
						debug!(
							"Root domain {} is allowed by url_preview_domain_explicit_allowlist (check 1/3)",
							&root_domain
						);
						return true;
					}

					if allowlist_domain_contains.iter().any(|domain_s| domain_s.contains(&root_domain.to_owned())) {
						debug!(
							"Root domain {} is allowed by url_preview_domain_contains_allowlist (check 2/3)",
							&root_domain
						);
						return true;
					}
				},
			}
		}
	}

	false
}