1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
//!
//! Implements rate limiting with governor.
//!
//! Would love to use tower_governor but I can't type it properly due to
//! https://github.com/benwis/tower-governor/issues/49.
//!
//! So we implement our own rate limiter here.
//!
use axum::response::{IntoResponse, Response};
use axum::{
body::Body,
http::{Request, StatusCode},
};
use futures_util::future::BoxFuture;
use governor::clock::QuantaClock;
use governor::state::keyed::DashMapStateStore;
use std::num::NonZero;
use std::sync::Arc;
use std::time::Duration;
use std::{convert::Infallible, task::Poll};
use tower::{Layer, Service};
use crate::core::extractors::PubkyHost;
use crate::quota_config::{LimitKey, LimitKeyType, PathLimit, RateUnit};
use crate::shared::HttpError;
use futures_util::StreamExt;
use governor::{Jitter, Quota, RateLimiter};
use super::extract_ip::extract_ip;
/// A Tower Layer to handle general rate limiting.
///
/// Supports rate limiting by request count and by upload/download speed.
///
/// Requires a `PubkyHostLayer` to be applied first.
/// Used to extract the user pubkey as the key for the rate limiter.
///
/// Returns 400 BAD REQUEST if the user pubkey aka pubky-host cannot be extracted.
///
#[derive(Debug, Clone)]
pub struct RateLimiterLayer {
limits: Vec<PathLimit>,
}
impl RateLimiterLayer {
/// Create a new rate limiter layer with the given quota.
///
/// If quota is None, rate limiting is disabled.
pub fn new(limits: Vec<PathLimit>) -> Self {
if limits.is_empty() {
tracing::info!("Rate limiting is disabled.");
} else {
let limits_str = limits
.iter()
.map(|limit| format!("\"{limit}\""))
.collect::<Vec<String>>();
tracing::info!("Rate limits configured: {}", limits_str.join(", "));
}
Self { limits }
}
}
impl<S> Layer<S> for RateLimiterLayer {
type Service = RateLimiterMiddleware<S>;
fn layer(&self, inner: S) -> Self::Service {
let tuples = self
.limits
.iter()
.map(|path| LimitTuple::new(path.clone()))
.collect();
RateLimiterMiddleware {
inner,
limits: tuples,
}
}
}
/// A tuple of a path limit and the actual governor rate limiter.
#[derive(Debug, Clone)]
struct LimitTuple {
pub limit: PathLimit,
pub limiter: Arc<RateLimiter<LimitKey, DashMapStateStore<LimitKey>, QuantaClock>>,
}
impl LimitTuple {
pub fn new(path_limit: PathLimit) -> Self {
let quota: Quota = path_limit.clone().into();
let limiter = Arc::new(RateLimiter::keyed(quota));
// Forget keys that are not used anymore. This is to prevent memory leaks.
let limiter_clone = limiter.clone();
tokio::spawn(async move {
let mut interval = tokio::time::interval(Duration::from_secs(60));
interval.tick().await;
loop {
interval.tick().await;
limiter_clone.retain_recent();
limiter_clone.shrink_to_fit();
}
});
Self {
limit: path_limit,
limiter,
}
}
/// Extract the key from the request.
///
/// The key is either the ip address of the client
/// or the user pubkey.
fn extract_key(&self, req: &Request<Body>) -> anyhow::Result<LimitKey> {
match self.limit.key {
LimitKeyType::Ip => extract_ip(req).map(LimitKey::Ip),
LimitKeyType::User => {
// Extract the user pubkey from the request.
req.extensions()
.get::<PubkyHost>()
.map(|pk| LimitKey::User(pk.public_key().clone()))
.ok_or(anyhow::anyhow!("Failed to extract user pubkey."))
}
}
}
/// Check if the request matches the limit.
pub fn is_match(&self, req: &Request<Body>) -> bool {
let path = req.uri().path();
let glob_match = self.limit.path.is_match(path);
let method_match = self.limit.method.0 == req.method();
glob_match && method_match
}
}
#[derive(Debug, Clone)]
pub struct RateLimiterMiddleware<S> {
inner: S,
limits: Vec<LimitTuple>,
}
impl<S> RateLimiterMiddleware<S> {
/// Throttle the upload body.
fn throttle_upload(
req: Request<Body>,
key: &LimitKey,
limiter: &Arc<RateLimiter<LimitKey, DashMapStateStore<LimitKey>, QuantaClock>>,
) -> Request<Body> {
let (parts, body) = req.into_parts();
let new_body = Self::throttle_body(body, key, limiter);
Request::from_parts(parts, new_body)
}
/// Throttle the download body.
fn throttle_download(
res: Response<Body>,
key: &LimitKey,
limiter: &Arc<RateLimiter<LimitKey, DashMapStateStore<LimitKey>, QuantaClock>>,
) -> Response<Body> {
let (parts, body) = res.into_parts();
let new_body = Self::throttle_body(body, key, limiter);
Response::from_parts(parts, new_body)
}
/// Throttle the up or download body.
///
/// Important: The speed quotas are always in kilobytes, not bytes.
/// Counting bytes is not practical.
///
fn throttle_body(
body: Body,
key: &LimitKey,
limiter: &Arc<RateLimiter<LimitKey, DashMapStateStore<LimitKey>, QuantaClock>>,
) -> Body {
let body_stream = body.into_data_stream();
let limiter = limiter.clone();
let key = key.clone();
let throttled = body_stream
.map(move |chunk| {
let limiter = limiter.clone();
let key = key.clone();
// When the rate limit is exceeded, we wait between 25ms and 500ms before retrying.
// This is to avoid overwhelming the server with requests when the rate limit is exceeded.
// Randomization is used to avoid thundering herd problem.
let jitter = Jitter::new(
Duration::from_millis(25),
Duration::from_millis(500),
);
async move {
let bytes = match chunk {
Ok(actual_chunk) => {
actual_chunk
}
Err(e) => return Err(e),
};
// --- Round up to the nearest kilobyte. ---
// Important: If the chunk is < 1KB, it will be rounded up to 1 kb.
// Many small uploads will be counted as more than they actually are.
// I am not too concerned about this though because small random disk writes are stressing
// the disk more anyway compared to larger writes.
// Why are we doing this? governor::Quota is defined as a u32. u32 can only count up to 4GB.
// To support 4GB/s+ limits we need to count in kilobytes.
//
// --- Chunk Size ---
// The chunk size is determined by the client library.
// Common chunk sizes: 16KB to 10MB.
// HTTP based uploads are usually between 256KB and 1MB.
// Asking the limiter for 1KB packets is tradeoff between
// - Not calling the limiter too much
// - Guaranteeing the call size (1kb) is low enough to not cause race condition issues.
let chunk_kilobytes = bytes.len().div_ceil(1024);
for _ in 0..chunk_kilobytes {
// Check each kilobyte
if limiter
.until_key_n_ready_with_jitter(
&key,
NonZero::new(1).expect("1 is always non zero"),
jitter,
)
.await.is_err()
{
// Requested rate (1kb) is higher then the set limit (x kb/s).
// This should never happen.
unreachable!("Rate limiting is based on the number of kilobytes, not bytes. So 1 kb should always be allowed.");
};
}
Ok(bytes)
}
})
.buffered(1);
Body::from_stream(throttled)
}
/// Get the limits that match the request.
fn get_limit_matches(&self, req: &Request<Body>) -> Vec<&LimitTuple> {
self.limits
.iter()
.filter(|limit| limit.is_match(req))
.collect()
}
}
impl<S> Service<Request<Body>> for RateLimiterMiddleware<S>
where
S: Service<Request<Body>, Response = axum::response::Response, Error = Infallible>
+ Send
+ 'static
+ Clone,
S::Future: Send + 'static,
{
type Response = S::Response;
type Error = S::Error;
type Future = BoxFuture<'static, Result<Self::Response, Self::Error>>;
fn poll_ready(&mut self, cx: &mut std::task::Context<'_>) -> Poll<Result<(), Self::Error>> {
self.inner.poll_ready(cx).map_err(|_| unreachable!()) // `Infallible` conversion
}
fn call(&mut self, mut req: Request<Body>) -> Self::Future {
let mut inner = self.inner.clone();
// Match the request path + method to the defined limits.
let limits = self.get_limit_matches(&req);
if limits.is_empty() {
// No limits matched, so we can just call the next layer.
return Box::pin(async move { inner.call(req).await.map_err(|_| unreachable!()) });
}
// Go through all the limits and check if we need to throttle or reject the request.
for limit in limits.clone() {
let key = match limit.extract_key(&req) {
Ok(key) => key,
Err(e) => {
// Failed to extract the key, so we reject the request.
// This should only happen if the pubky-host couldn't be extracted.
tracing::warn!(
"{} {} Failed to extract key for rate limiting: {}",
limit.limit.path.0,
limit.limit.method.0,
e
);
return Box::pin(async move {
Ok(HttpError::new_with_message(
StatusCode::BAD_REQUEST,
"Failed to extract key for rate limiting",
)
.into_response())
});
}
};
if limit.limit.is_whitelisted(&key) {
continue;
}
match limit.limit.quota.rate_unit {
RateUnit::SpeedRateUnit(_) => {
// Speed limiting is enabled, so we need to throttle the upload.
req = Self::throttle_upload(req, &key, &limit.limiter);
}
RateUnit::Request => {
// Request limiting is enabled, so we need to limit the number of requests.
if let Err(e) = limit.limiter.check_key(&key) {
tracing::debug!(
"Rate limit of {} exceeded for {key}: {}",
limit.limit.quota,
e
);
return Box::pin(async move {
Ok(HttpError::new_with_message(
StatusCode::TOO_MANY_REQUESTS,
"Rate limit exceeded",
)
.into_response())
});
};
}
};
}
// Create a clone of the request without the body.
// This way, we can extract the keys for the response too.
let (parts, body) = req.into_parts();
let req_clone = Request::from_parts(parts.clone(), Body::empty());
let req = Request::from_parts(parts, body);
let speed_limits = limits
.into_iter()
.filter(|limit| limit.limit.quota.rate_unit.is_speed_rate_unit())
.cloned()
.collect::<Vec<_>>();
Box::pin(async move {
// Call the next layer and receive the response.
let mut response = match inner.call(req).await.map_err(|_| unreachable!()) {
Ok(response) => response,
Err(e) => return Err(e),
};
// Rate limit the download speed.
for limit in speed_limits {
if let Ok(key) = limit.extract_key(&req_clone) {
response = Self::throttle_download(response, &key, &limit.limiter);
};
}
Ok(response)
})
}
}
#[cfg(test)]
mod tests {
use std::net::{IpAddr, Ipv4Addr, SocketAddr};
use axum::http::Method;
use axum::{
routing::{get, post},
Router,
};
use axum_server::Server;
use pkarr::{Keypair, PublicKey};
use reqwest::{Client, Response};
use tokio::{task::JoinHandle, time::Instant};
use crate::shared::HttpResult;
use crate::{core::layers::pubky_host::PubkyHostLayer, quota_config::GlobPattern};
use super::*;
// Fake upload handler that just consumes the body.
pub async fn upload_handler(body: Body) -> HttpResult<impl IntoResponse> {
let mut stream = body.into_data_stream();
while let Some(chunk) = stream.next().await.transpose()? {
// Consume body
let _ = chunk;
}
Ok((StatusCode::CREATED, ()))
}
// Fake upload handler that just consumes the body.
pub async fn download_handler() -> HttpResult<impl IntoResponse> {
let response_body = vec![0u8; 3 * 1024]; // 3kb
Ok((StatusCode::OK, response_body))
}
// Start a server with the given quota config on a random port.
async fn start_server(config: Vec<PathLimit>) -> SocketAddr {
let app = Router::new()
.route("/upload", post(upload_handler))
.route("/download", get(download_handler))
.layer(RateLimiterLayer::new(config))
.layer(PubkyHostLayer);
// Create a TCP listener to bind to the socket first
// Use port 0 to let the OS assign a random available port
let listener = tokio::net::TcpListener::bind(SocketAddr::new(
IpAddr::V4(Ipv4Addr::new(127, 0, 0, 1)),
0,
))
.await
.unwrap();
// Get the actual socket address with the OS-assigned port
let socket = listener.local_addr().unwrap();
// Use the listener with axum_server
let server = Server::from_tcp(listener.into_std().unwrap());
tokio::spawn(async move {
server
.serve(app.into_make_service_with_connect_info::<SocketAddr>())
.await
.unwrap();
});
socket
}
#[tokio::test]
async fn test_throttle_upload() {
let path_limit = PathLimit::new(
GlobPattern::new("/upload"),
Method::POST,
"1kb/s".parse().unwrap(),
LimitKeyType::Ip,
None,
);
let socket = start_server(vec![path_limit]).await;
fn upload_data(socket: SocketAddr, kilobytes: usize) -> JoinHandle<()> {
tokio::spawn(async move {
let client = Client::new();
let data = vec![0u8; kilobytes * 1024];
let response = client
.post(format!("http://{}/upload", socket))
.body(data.clone())
.send()
.await
.unwrap();
assert_eq!(response.status(), StatusCode::CREATED);
})
}
let start = Instant::now();
// Spawn in the background to test 2 uploads in parallel
// Upload 3kb each
let handle1 = upload_data(socket, 4);
let handle2 = upload_data(socket, 4);
// Wait for the uploads to finish
let _ = tokio::try_join!(handle1, handle2);
let time_taken = start.elapsed();
assert!(time_taken > Duration::from_secs(6), "Should at least take 6s because uploads are limited to 1kb/s and the sum of the uploads is 6kb");
}
#[tokio::test]
async fn test_throttle_download() {
let path_limit = PathLimit::new(
GlobPattern::new("/download"),
Method::GET,
"1kb/s".parse().unwrap(),
LimitKeyType::Ip,
None,
);
let socket = start_server(vec![path_limit]).await;
fn download_data(socket: SocketAddr) -> JoinHandle<()> {
tokio::spawn(async move {
let client = Client::new();
let response = client
.get(format!("http://{}/download", socket))
.send()
.await
.unwrap();
assert_eq!(response.status(), StatusCode::OK);
response.bytes().await.unwrap(); // Download the body
})
}
let start = Instant::now();
// Spawn in the background to test 2 downloads in parallel
// Download 3kb each
let handle1 = download_data(socket);
let handle2 = download_data(socket);
// Wait for the uploads to finish
let _ = tokio::try_join!(handle1, handle2);
let time_taken = start.elapsed();
if time_taken < Duration::from_secs(6) {
panic!("Should at least take 6s because downloads are limited to 1kb/s and the sum of the downloads is 6kb. Time taken: {:?}", time_taken);
}
}
#[tokio::test]
async fn test_limit_parallel_requests_with_ip_key() {
let path_limit = PathLimit::new(
GlobPattern::new("/upload"),
Method::POST,
"1r/m".parse().unwrap(),
LimitKeyType::Ip,
None,
);
let socket = start_server(vec![path_limit]).await;
fn send_request(socket: SocketAddr) -> JoinHandle<Response> {
tokio::spawn(async move {
let client = Client::new();
let response = client
.post(format!("http://{}/upload", socket))
.send()
.await
.unwrap();
response
})
}
// Spawn in the background to test 2 uploads in parallel
let handle1 = send_request(socket);
let handle2 = send_request(socket);
// Wait for the uploads to finish
let (res1, res2) = tokio::try_join!(handle1, handle2).unwrap();
assert_eq!(res1.status(), StatusCode::CREATED);
assert_eq!(res2.status(), StatusCode::TOO_MANY_REQUESTS);
}
#[tokio::test]
async fn test_limit_parallel_requests_with_user_key() {
let path_limit = PathLimit::new(
GlobPattern::new("/upload"),
Method::POST,
"1r/m".parse().unwrap(),
LimitKeyType::User,
None,
);
let socket = start_server(vec![path_limit]).await;
fn send_request(socket: SocketAddr, user_pubkey: PublicKey) -> JoinHandle<Response> {
tokio::spawn(async move {
let client = Client::new();
let response = client
.post(format!("http://{}/upload?pubky-host={user_pubkey}", socket))
.send()
.await
.unwrap();
response
})
}
// Spawn in the background to test 2 uploads in parallel
let user1_pubkey = Keypair::random().public_key();
let handle1 = send_request(socket, user1_pubkey.clone());
let handle2 = send_request(socket, user1_pubkey.clone());
let user2_pubkey = Keypair::random().public_key();
let handle3 = send_request(socket, user2_pubkey.clone());
// Wait for the uploads to finish
let (res1, res2, res3) = tokio::try_join!(handle1, handle2, handle3).unwrap();
assert_eq!(res1.status(), StatusCode::CREATED);
assert_eq!(res2.status(), StatusCode::TOO_MANY_REQUESTS);
assert_eq!(res3.status(), StatusCode::CREATED);
}
}