* v5: Health Check & LeastLoad Strategy (rebased from 2c5a71490368500a982018a74a6d519c7e121816)
Some changes will be necessary to integrate it into V2Ray
* Update proto
* parse duration conf with time.Parse()
* moving health ping to observatory as a standalone component
* moving health ping to observatory as a standalone component: auto generated file
* add initialization for health ping
* incorporate changes in router implementation
* support principle target output
* add v4 json support for BurstObservatory & fix balancer reference
* update API command
* remove cancelled API
* return zero length value when observer is not found
* remove duplicated targeted dispatch
* adjust test with updated structure
* bug fix for observer
* fix strategy selector
* fix strategy least load
* Fix ticker usage
ticker.Close does not close ticker.C
* feat: Replace default Health Ping URL to HTTPS (#1991)
* fix selectLeastLoad() returns wrong number of nodes (#2083)
* Test: fix leastload strategy unit test
* fix(router): panic caused by concurrent map read and write (#2678)
* Clean up code
---------
Co-authored-by: Jebbs <qjebbs@gmail.com>
Co-authored-by: Shelikhoo <xiaokangwang@outlook.com>
Co-authored-by: 世界 <i@sekai.icu>
Co-authored-by: Bernd Eichelberger <46166740+4-FLOSS-Free-Libre-Open-Source-Software@users.noreply.github.com>
Co-authored-by: 秋のかえで <autmaple@protonmail.com>
Co-authored-by: Rinka <kujourinka@gmail.com>
Android client prepares an IP before proxy connection is established. It is useful when connecting to wireguard (or quic) outbound with domain address. E.g. engage.cloudflareclient.com:2408
* allow set interface under windows
Signed-off-by: San Zhang <zhangan@mail.com>
* polish code
Signed-off-by: San Zhang <zhangan@mail.com>
---------
Signed-off-by: San Zhang <zhangan@mail.com>
Co-authored-by: San Zhang <zhangan@mail.com>
Issue #2605 brought up real problem that QUIC dialer doesn't support sockopt at the moment. Inside `internet.DialSystem(...)` function, one of the branch that involves `redirect(...)` returns `cnc.connection` instance that is currently unhandled by the code logic, and thus caused program panic during runtime.
It seems the sockopt support for QUIC protocol requires a couple changes including making `cnc.connection` public, such that we can handle in dialer, along with some thorough tests, this commit simply adds safety check to explicity state the fact that QUIC isn't working with sockopt. And the implementation of the feature can be scheduled later on.
* Added tcp fragmentation for freedom outbound
* Added TCP_NODELAY to outbound sockopt
* Changed fragment parameters to accept ranges and changed strategy to use length
* Changed packetNumber to packets, supporting range.
* Refactored the freedom fragment logic
* Refine Write()
---------
Co-authored-by: RPRX <63339210+RPRX@users.noreply.github.com>
* Add fingerprint xray_random
xray_random means to pick a random uTLS fingerprint at the core startup
This way, the fingerprint is stable for a user for some days. While there is no identifiable signature for the whole xray community
* Fingerprint "random" refine
Exclude old fingerprint from RNG
* DialSystem for Quic
DialSystem() is needed in case of Android client,
where the raw conn is protected for vpn service
* Fix client dialer log
Log such as:
tunneling request to tcp:www.google.com:80 via tcp:x.x.x.x:443
the second "tcp" is misleading when using mKcp or quic transport
Remove the second "tcp" and add the correct logging for transport dialer:
- transport/internet/tcp: dialing TCP to tcp:x.x.x.x:443
- transport/internet/quic: dialing quic to udp:x.x.x.x:443
* Quic new stream allocation mode
Currently this is how Quic works: client muxing all tcp and udp traffic through a single session, when there are more than 32 running streams in the session,
the next stream request will fail and open with a new session (port). Imagine lineup the session from left to right:
|
| |
| | |
As the streams finishes, we still open stream from the left, original session. So the base session will always be there and new sessions on the right come and go.
However, either due to QOS or bugs in Quic implementation, the traffic "wear out" the base session. It will become slower and in the end not receiving any data from server side.
I couldn't figure out a solution for this problem at the moment, as a workaround:
| |
| | |
| | |
I came up with this new stream allocation mode, that it will never open new streams in the old sessions, but only from current or new session from right.
The keeplive config is turned off from server and client side. This way old sessions will natually close and new sessions keep generating.
Note the frequency of new session is still controlled by the server side. Server can assign a large max stream limit. In this case the new allocation mode will be similar to the current mode.
* Increase some tls test timeout
* Fix TestUserValidator
* Change all tests to VMessAEAD
Old VMess MD5 tests will be rejected and fail in 2022
* Chore: auto format code
* protocol harmonization with V2Ray/V2Fly by supporting both V2Ray server and XRay server
* protocol harmonization with V2Ray/V2Fly by supporting both V2Ray server and XRay server comment
* verify peer cert function for better man in the middle prevention
* publish cert chain hash generation algorithm
* added calculation of certificate hash as separate command and tlsping, use base64 to represent fingerprint to align with jsonPb
* apply coding style
* added test case for pinned certificates
* refactored cert pin
* pinned cert test
* added json loading of the PinnedPeerCertificateChainSha256
* removed tool to prepare for v5
* Add server cert pinning for Xtls
Change command "xray tls certChainHash" to xray style
Co-authored-by: Shelikhoo <xiaokangwang@outlook.com>
When gRPC transport have been configured to use TLS, it may silently ignore TLS failure. This may make it harder to diagnose TLS setting issues when gRPC transport is used. This comment is added to help other developers be aware of this caveat.
Co-authored-by: Shelikhoo <xiaokangwang@outlook.com>
