1
0
Fork 0
mirror of https://gitlab.com/famedly/conduit.git synced 2025-08-11 17:50:59 +00:00
conduit/src/api/client_server/media.rs

517 lines
17 KiB
Rust
Raw Normal View History

2023-06-25 19:31:40 +02:00
use std::time::Duration;
use crate::{service::media::{FileMeta, UrlPreviewData}, services, utils, Error, Result, Ruma};
2020-07-30 18:14:47 +02:00
use ruma::api::client::{
2023-07-28 17:53:15 +02:00
error::{ErrorKind, RetryAfter},
2022-02-18 15:33:14 +01:00
media::{
create_content, get_content, get_content_as_filename, get_content_thumbnail,
2023-07-28 17:53:15 +02:00
get_media_config, get_media_preview
},
2020-07-30 18:14:47 +02:00
};
2023-07-28 17:53:15 +02:00
use {
webpage::HTML,
reqwest::Url,
std::{io::Cursor, net::IpAddr, sync::Arc},
2023-07-28 17:53:15 +02:00
tokio::sync::Notify,
image::io::Reader as ImgReader,
};
const MXC_LENGTH: usize = 32;
2020-07-30 18:14:47 +02:00
2021-08-31 19:14:37 +02:00
/// # `GET /_matrix/media/r0/config`
///
/// Returns max upload size.
pub async fn get_media_config_route(
2022-02-18 15:33:14 +01:00
_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(),
})
2020-07-30 18:14:47 +02:00
}
2023-07-28 17:53:15 +02:00
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 max_download_size = 300_000;
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() > max_download_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 {
// 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.
2023-07-28 17:53:15 +02:00
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()
2023-07-28 17:53:15 +02:00
|| 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)
2023-07-28 17:53:15 +02:00
|| 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 b >= 0x20 && b <= 0x2F)
))
|| ((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
2023-07-28 17:53:15 +02:00
}
}
}
async fn request_url_preview(url: &str) -> Result<UrlPreviewData> {
2023-07-28 17:53:15 +02:00
let client = services().globals.default_client();
let response = client.head(url).send().await?;
2023-07-28 17:53:15 +02:00
if !response
.remote_addr()
.map_or(false, |a| url_request_allowed(&a.ip()))
{
return Err(Error::BadRequest(
ErrorKind::Unknown,
"Requesting from this address 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?,
2023-07-28 17:53:15 +02:00
_ => {
return Err(Error::BadRequest(
ErrorKind::Unknown,
"Unsupported Content-Type",
))
}
};
services().media.set_url_preview(url, &data).await?;
2023-07-28 17:53:15 +02:00
Ok(data)
}
async fn get_url_preview(url: &str) -> Result<UrlPreviewData> {
if let Some(preview) = services().media.get_url_preview(url).await {
2023-07-28 17:53:15 +02:00
return Ok(preview);
}
let notif_opt = services()
.media
.url_preview_requests
.read()
.unwrap()
.get(url)
2023-07-28 17:53:15 +02:00
.cloned();
match notif_opt {
None => {
let notifier = Arc::new(Notify::new());
{
services().media
.url_preview_requests
.write()
.unwrap()
.insert(url.to_owned(), notifier.clone());
2023-07-28 17:53:15 +02:00
}
let data = request_url_preview(url).await;
2023-07-28 17:53:15 +02:00
notifier.notify_waiters();
{
services().media.url_preview_requests.write().unwrap().remove(url);
2023-07-28 17:53:15 +02:00
}
data
}
Some(notifier) => {
// wait until being notified that request is finished
let notifier = notifier.clone();
let notifier = notifier.notified();
notifier.await;
services().media
.get_url_preview(url)
2023-07-28 17:53:15 +02:00
.await
.ok_or(Error::BadRequest(
ErrorKind::Unknown,
"No Preview available",
))
}
}
}
fn url_preview_allowed(url_str: &str) -> bool {
const DEFAULT_ALLOWLIST: &[&str] = &[
"matrix.org",
"mastodon.social",
"youtube.com",
"wikipedia.org",
];
let url = match Url::parse(url_str) {
Ok(u) => u,
Err(_) => return false,
};
if ["http", "https"].iter().all(|&scheme| scheme != url.scheme().to_lowercase()) {
return false;
}
let mut host = match url.host_str() {
None => return false,
Some(h) => h.to_lowercase(),
};
let allowlist = services().globals.url_preview_allowlist();
if allowlist.contains(&"*".to_owned()) {
return true;
}
while !host.is_empty() {
if allowlist.contains(&host) {
return true;
}
if allowlist.contains(&"default".to_owned()) && DEFAULT_ALLOWLIST.contains(&host.as_str()) {
return true;
}
/* also check higher level domains, so that e.g. `en.m.wikipedia.org` is matched by `wikipedia.org` on allowlist. */
host = match host.split_once('.') {
None => return false,
Some((_, domain)) => domain.to_owned(),
}
}
false
}
2023-07-28 17:53:15 +02:00
/// # `GET /_matrix/media/r0/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) {
2023-07-28 17:53:15 +02:00
return Err(Error::BadRequest(
ErrorKind::Unknown,
"Previewing URL not allowed",
));
}
if let Ok(preview) = get_url_preview(url).await {
2023-07-28 17:53:15 +02:00
let res = serde_json::value::to_raw_value(&preview).expect("Converting to JSON failed");
return Ok(get_media_preview::v3::Response::from_raw_value(res));
}
Err(Error::BadRequest(
ErrorKind::LimitExceeded {
retry_after: Some(RetryAfter::Delay(Duration::from_secs(5))),
},
"Retry later",
))
}
2021-08-31 19:14:37 +02:00
/// # `POST /_matrix/media/r0/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(
2022-12-14 13:09:10 +01:00
body: Ruma<create_content::v3::Request>,
2022-02-18 15:33:14 +01:00
) -> Result<create_content::v3::Response> {
2020-07-30 18:14:47 +02:00
let mxc = format!(
"mxc://{}/{}",
services().globals.server_name(),
2020-07-30 18:14:47 +02:00
utils::random_string(MXC_LENGTH)
);
2021-06-06 16:58:32 +04:30
2022-10-05 20:34:31 +02:00
services()
.media
2021-06-06 16:58:32 +04:30
.create(
mxc.clone(),
2022-10-05 20:34:31 +02:00
body.filename
2021-06-06 16:58:32 +04:30
.as_ref()
.map(|filename| "inline; filename=".to_owned() + filename)
.as_deref(),
2022-10-05 15:33:57 +02:00
body.content_type.as_deref(),
2021-06-06 16:58:32 +04:30
&body.file,
)
.await?;
2020-07-30 18:14:47 +02:00
2022-02-18 15:33:14 +01:00
Ok(create_content::v3::Response {
content_uri: mxc.into(),
2020-12-19 16:00:11 +01:00
blurhash: None,
})
2020-07-30 18:14:47 +02:00
}
pub async fn get_remote_content(
mxc: &str,
server_name: &ruma::ServerName,
2022-12-14 13:09:10 +01:00
media_id: String,
2022-02-18 15:33:14 +01:00
) -> Result<get_content::v3::Response, Error> {
let content_response = services()
.sending
.send_federation_request(
server_name,
2022-02-18 15:33:14 +01:00
get_content::v3::Request {
allow_remote: false,
2022-12-14 13:09:10 +01:00
server_name: server_name.to_owned(),
2022-01-27 17:08:04 +01:00
media_id,
2023-06-25 19:31:40 +02:00
timeout_ms: Duration::from_secs(20),
allow_redirect: false,
},
)
.await?;
2022-10-05 20:34:31 +02:00
services()
.media
.create(
2022-11-21 09:51:39 +01:00
mxc.to_owned(),
2022-10-05 15:33:57 +02:00
content_response.content_disposition.as_deref(),
content_response.content_type.as_deref(),
&content_response.file,
)
.await?;
Ok(content_response)
}
/// # `GET /_matrix/media/r0/download/{serverName}/{mediaId}`
2021-08-31 19:14:37 +02:00
///
/// Load media from our server or over federation.
///
/// - Only allows federation if `allow_remote` is true
pub async fn get_content_route(
2022-12-14 13:09:10 +01:00
body: Ruma<get_content::v3::Request>,
2022-02-18 15:33:14 +01:00
) -> Result<get_content::v3::Response> {
let mxc = format!("mxc://{}/{}", body.server_name, body.media_id);
2020-07-30 18:14:47 +02:00
if let Some(FileMeta {
content_disposition,
content_type,
2020-07-30 18:14:47 +02:00
file,
2022-09-07 13:25:51 +02:00
}) = services().media.get(mxc.clone()).await?
2020-07-30 18:14:47 +02:00
{
2022-02-18 15:33:14 +01:00
Ok(get_content::v3::Response {
2020-07-30 18:14:47 +02:00
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 {
2022-01-27 17:08:04 +01:00
let remote_content_response =
2022-12-14 13:09:10 +01:00
get_remote_content(&mxc, &body.server_name, body.media_id.clone()).await?;
2024-05-05 09:35:49 +02:00
Ok(get_content::v3::Response {
content_disposition: remote_content_response.content_disposition,
content_type: remote_content_response.content_type,
2024-05-05 09:35:49 +02:00
file: remote_content_response.file,
cross_origin_resource_policy: Some("cross-origin".to_owned()),
})
2020-07-30 18:14:47 +02:00
} else {
Err(Error::BadRequest(ErrorKind::NotFound, "Media not found."))
}
}
/// # `GET /_matrix/media/r0/download/{serverName}/{mediaId}/{fileName}`
///
/// Load media from our server or over federation, permitting desired filename.
///
/// - Only allows federation if `allow_remote` is true
pub async fn get_content_as_filename_route(
2022-12-14 13:09:10 +01:00
body: Ruma<get_content_as_filename::v3::Request>,
2022-02-18 15:33:14 +01:00
) -> Result<get_content_as_filename::v3::Response> {
let mxc = format!("mxc://{}/{}", body.server_name, body.media_id);
if let Some(FileMeta {
file, content_type, ..
}) = services().media.get(mxc.clone()).await?
{
2022-02-18 15:33:14 +01:00
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 {
2022-01-27 17:08:04 +01:00
let remote_content_response =
2022-12-14 13:09:10 +01:00
get_remote_content(&mxc, &body.server_name, body.media_id.clone()).await?;
2022-02-18 15:33:14 +01:00
Ok(get_content_as_filename::v3::Response {
content_disposition: Some(format!("inline: filename={}", body.filename)),
content_type: remote_content_response.content_type,
2022-01-27 17:08:04 +01:00
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/r0/thumbnail/{serverName}/{mediaId}`
2021-08-31 19:14:37 +02:00
///
/// Load media thumbnail from our server or over federation.
///
/// - Only allows federation if `allow_remote` is true
pub async fn get_content_thumbnail_route(
2022-12-14 13:09:10 +01:00
body: Ruma<get_content_thumbnail::v3::Request>,
2022-02-18 15:33:14 +01:00
) -> Result<get_content_thumbnail::v3::Response> {
let mxc = format!("mxc://{}/{}", body.server_name, body.media_id);
2020-07-30 18:14:47 +02:00
if let Some(FileMeta {
2024-05-05 09:35:49 +02:00
file, content_type, ..
}) = services()
2021-06-06 16:58:32 +04:30
.media
.get_thumbnail(
2022-09-07 13:25:51 +02:00
mxc.clone(),
2021-06-06 16:58:32 +04:30
body.width
.try_into()
.map_err(|_| Error::BadRequest(ErrorKind::InvalidParam, "Width is invalid."))?,
body.height
.try_into()
.map_err(|_| Error::BadRequest(ErrorKind::InvalidParam, "Width 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 {
let get_thumbnail_response = services()
2020-12-19 16:00:11 +01:00
.sending
.send_federation_request(
&body.server_name,
2022-02-18 15:33:14 +01:00
get_content_thumbnail::v3::Request {
2020-12-19 16:00:11 +01:00
allow_remote: false,
height: body.height,
width: body.width,
method: body.method.clone(),
2022-12-14 13:09:10 +01:00
server_name: body.server_name.clone(),
media_id: body.media_id.clone(),
2023-06-25 19:31:40 +02:00
timeout_ms: Duration::from_secs(20),
allow_redirect: false,
2020-12-19 16:00:11 +01:00
},
)
.await?;
2022-10-05 20:34:31 +02:00
services()
.media
2021-06-06 16:58:32 +04:30
.upload_thumbnail(
mxc,
2022-10-05 15:33:57 +02:00
None,
get_thumbnail_response.content_type.as_deref(),
2021-06-06 16:58:32 +04:30
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)
2020-07-30 18:14:47 +02:00
} else {
Err(Error::BadRequest(ErrorKind::NotFound, "Media not found."))
}
}