| Message ID | cover.1725935420.git.lucien.xin@gmail.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | net: implement the QUIC protocol in linux kernel | expand |
On Mon, Sep 09, 2024 at 10:30:15PM -0400, Xin Long wrote: > 4. Performance testing via iperf > > The performance testing was conducted using iperf [5] over a 100G > physical NIC, evaluating various packet sizes and MTUs: > > - QUIC vs. kTLS: > > UNIT size:1024 size:4096 size:16384 size:65536 > Gbits/sec QUIC | kTLS QUIC | kTLS QUIC | kTLS QUIC | kTLS > ──────────────────────────────────────────────────────────────────── > mtu:1500 1.67 | 2.16 3.04 | 5.04 3.49 | 7.84 3.83 | 7.95 > ──────────────────────────────────────────────────────────────────── > mtu:9000 2.17 | 2.41 5.47 | 6.19 6.45 | 8.66 7.48 | 8.90 > > - QUIC(disable_1rtt_encryption) vs. TCP: > > UNIT size:1024 size:4096 size:16384 size:65536 > Gbits/sec QUIC | TCP QUIC | TCP QUIC | TCP QUIC | TCP > ──────────────────────────────────────────────────────────────────── > mtu:1500 2.17 | 2.49 3.59 | 8.36 6.09 | 15.1 6.92 | 16.2 > ──────────────────────────────────────────────────────────────────── > mtu:9000 2.47 | 2.54 7.66 | 7.97 14.7 | 20.3 19.1 | 31.3 > > > The performance gap between QUIC and kTLS may be attributed to: > > - The absence of Generic Segmentation Offload (GSO) for QUIC. > - An additional data copy on the transmission (TX) path. > - Extra encryption required for header protection in QUIC. > - A longer header length for the stream data in QUIC. > This is not appealing. However, I can offer you one more possible advantage of in-kernel QUIC. You can think about adding iouring support for QUIC socket, because that could possibly chain the socket fastpath operations together which opens the door for more optimization. Just my two cents. Thanks!
On Fri, Sep 13, 2024 at 9:05 PM Cong Wang <xiyou.wangcong@gmail.com> wrote: > > On Mon, Sep 09, 2024 at 10:30:15PM -0400, Xin Long wrote: > > 4. Performance testing via iperf > > > > The performance testing was conducted using iperf [5] over a 100G > > physical NIC, evaluating various packet sizes and MTUs: > > > > - QUIC vs. kTLS: > > > > UNIT size:1024 size:4096 size:16384 size:65536 > > Gbits/sec QUIC | kTLS QUIC | kTLS QUIC | kTLS QUIC | kTLS > > ──────────────────────────────────────────────────────────────────── > > mtu:1500 1.67 | 2.16 3.04 | 5.04 3.49 | 7.84 3.83 | 7.95 > > ──────────────────────────────────────────────────────────────────── > > mtu:9000 2.17 | 2.41 5.47 | 6.19 6.45 | 8.66 7.48 | 8.90 > > > > - QUIC(disable_1rtt_encryption) vs. TCP: > > > > UNIT size:1024 size:4096 size:16384 size:65536 > > Gbits/sec QUIC | TCP QUIC | TCP QUIC | TCP QUIC | TCP > > ──────────────────────────────────────────────────────────────────── > > mtu:1500 2.17 | 2.49 3.59 | 8.36 6.09 | 15.1 6.92 | 16.2 > > ──────────────────────────────────────────────────────────────────── > > mtu:9000 2.47 | 2.54 7.66 | 7.97 14.7 | 20.3 19.1 | 31.3 > > > > > > The performance gap between QUIC and kTLS may be attributed to: > > > > - The absence of Generic Segmentation Offload (GSO) for QUIC. > > - An additional data copy on the transmission (TX) path. > > - Extra encryption required for header protection in QUIC. > > - A longer header length for the stream data in QUIC. > > > > This is not appealing. > > However, I can offer you one more possible advantage of in-kernel QUIC. > You can think about adding iouring support for QUIC socket, because that > could possibly chain the socket fastpath operations together which opens > the door for more optimization. > I haven't had the chance to try io_uring. From what I understand, it doesn’t require any changes to the protocol stack (Linux QUIC doesn’t support MSG_ZEROCOPY at this time). Thanks for offering this suggestion, it sounds worth exploring when using the QUIC socket in the future. Thanks.
Introduction ============ The QUIC protocol, as defined in RFC9000, offers a UDP-based, secure transport with flow-controlled streams for efficient communication, low-latency connection setup, and network path migration, ensuring confidentiality, integrity, and availability across various deployments. This implementation introduces QUIC support in Linux Kernel, offering several key advantages: - Seamless Integration for Kernel Subsystems: Kernel subsystems such as SMB and NFS can operate over QUIC seamlessly after the handshake, leveraging the net/handshake APIs. - Standardized Socket APIs for QUIC: This implementation standardizes the socket APIs for QUIC, covering essential operations like listen, accept, connect, sendmsg, recvmsg, close, get/setsockopt, and getsock/peername(). - Efficient ALPN Routing: It incorporates ALPN routing within the kernel, efficiently directing incoming requests to the appropriate applications across different processes based on ALPN. - Performance Enhancements: By minimizing data duplication through zero-copy techniques such as sendfile(), and paving the way for crypto offloading in NICs, this implementation enhances performance and prepares for future optimizations. This implementation offers fundamental support for the following RFCs: - RFC9000 - QUIC: A UDP-Based Multiplexed and Secure Transport - RFC9001 - Using TLS to Secure QUIC - RFC9002 - QUIC Loss Detection and Congestion Control - RFC9221 - An Unreliable Datagram Extension to QUIC - RFC9287 - Greasing the QUIC Bit - RFC9368 - Compatible Version Negotiation for QUIC - RFC9369 - QUIC Version 2 The socket APIs for QUIC follow the RFC draft [1]: - The Sockets API Extensions for In-kernel QUIC Implementations Implementation ============== The core idea is to implement QUIC within the kernel, using a userspace handshake approach. Only the processing and creation of raw TLS Handshake Messages are handled in userspace, facilitated by a TLS library like GnuTLS. These messages are exchanged between kernel and userspace via sendmsg() and recvmsg(), with cryptographic details conveyed through control messages (cmsg). The entire QUIC protocol, aside from the TLS Handshake Messages processing and creation, is managed within the kernel. Rather than using a Upper Layer Protocol (ULP) layer, this implementation establishes a socket of type IPPROTO_QUIC (similar to IPPROTO_MPTCP), operating over UDP tunnels. Kernel consumers can initiate a handshake request from the kernel to userspace using the existing net/handshake netlink. The userspace component, such as tlshd service [2], then manages the processing of the QUIC handshake request. - Handshake Architecture: ┌──────┐ ┌──────┐ │ APP1 │ │ APP2 │ ... └──────┘ └──────┘ ┌──────────────────────────────────────────┐ │ {quic_client/server_handshake()} │<─────────────┐ └──────────────────────────────────────────┘ ┌─────────────┐ {send/recvmsg()} {set/getsockopt()} │ tlshd │ [CMSG handshake_info] [SOCKOPT_CRYPTO_SECRET] └─────────────┘ [SOCKOPT_TRANSPORT_PARAM_EXT] │ ^ │ ^ │ ^ │ │ Userspace │ │ │ │ │ │ ──────────────│─│──────────────────│─│──────────────────│───│──────── Kernel │ │ │ │ │ │ v │ v │ v │ ┌──────────────────────────────────────────┐ ┌─────────────┐ │ socket (IPPROTO_QUIC) | protocol │<──┐ │ handshake │ ├──────────────────────────────────────────┤ │ │netlink APIs │ │ stream | connid | cong | path | timer │ │ └─────────────┘ ├──────────────────────────────────────────┤ │ │ │ │ packet | frame | crypto | pnmap │ │ ┌─────┐ ┌─────┐ ├──────────────────────────────────────────┤ │ │ │ │ │ │ input | output │ │───│ SMB │ │ NFS │... ├──────────────────────────────────────────┤ │ │ │ │ │ │ UDP tunnels │ │ └─────┘ └─────┘ └──────────────────────────────────────────┘ └──────┴───────┘ - User Data Architecture: ┌──────┐ ┌──────┐ │ APP1 │ │ APP2 │ ... └──────┘ └──────┘ {send/recvmsg()} {set/getsockopt()} [CMSG stream_info] [SOCKOPT_KEY_UPDATE] [SOCKOPT_CONNECTION_MIGRATION] [SOCKOPT_STREAM_OPEN/RESET/STOP_SENDING] │ ^ │ ^ Userspace │ │ │ │ ──────────────│─│──────────────────│─│──────────────────────── Kernel │ │ │ │ v │ v │ ┌──────────────────────────────────────────┐ │ socket (IPPROTO_QUIC) | protocol │<──┐{kernel_send/recvmsg()} ├──────────────────────────────────────────┤ │{kernel_set/getsockopt()} │ stream | connid | cong | path | timer │ │ ├──────────────────────────────────────────┤ │ │ packet | frame | crypto | pnmap │ │ ┌─────┐ ┌─────┐ ├──────────────────────────────────────────┤ │ │ │ │ │ │ input | output │ │───│ SMB │ │ NFS │... ├──────────────────────────────────────────┤ │ │ │ │ │ │ UDP tunnels │ │ └─────┘ └─────┘ └──────────────────────────────────────────┘ └──────┴───────┘ Usage ===== This implementation supports a mapping of QUIC into sockets APIs. Similar to TCP and SCTP, a typical Server and Client use the following system call sequence to communicate: Client Server ────────────────────────────────────────────────────────────────────── sockfd = socket(IPPROTO_QUIC) listenfd = socket(IPPROTO_QUIC) bind(sockfd) bind(listenfd) listen(listenfd) connect(sockfd) quic_client_handshake(sockfd) sockfd = accecpt(listenfd) quic_server_handshake(sockfd, cert) sendmsg(sockfd) recvmsg(sockfd) close(sockfd) close(sockfd) close(listenfd) Please note that quic_client_handshake() and quic_server_handshake() functions are currently sourced from libquic [3]. These functions are responsible for receiving and processing the raw TLS handshake messages until the completion of the handshake process. For utilization by kernel consumers, it is essential to have tlshd service [2] installed and running in userspace. This service receives and manages kernel handshake requests for kernel sockets. In the kernel, the APIs closely resemble those used in userspace: Client Server ──────────────────────────────────────────────────────────────────────── __sock_create(IPPROTO_QUIC, &sock) __sock_create(IPPROTO_QUIC, &sock) kernel_bind(sock) kernel_bind(sock) kernel_listen(sock) kernel_connect(sock) tls_client_hello_x509(args:{sock}) kernel_accept(sock, &newsock) tls_server_hello_x509(args:{newsock}) kernel_sendmsg(sock) kernel_recvmsg(newsock) sock_release(sock) sock_release(newsock) sock_release(sock) Please be aware that tls_client_hello_x509() and tls_server_hello_x509() are APIs from net/handshake/. They are used to dispatch the handshake request to the userspace tlshd service and subsequently block until the handshake process is completed. Tests ===== 1. Functional testing The testing can be run by `make check` in libquic [3] and include: - [Function Tests (PSK)] - [Function Tests (Certificate)] - [Performance Tests (IPv4)] - [Performance Tests (IPv6, Disable 1RTT Encryption)] - [Performance Tests (IPv6)] - [Performance Tests (IPv4, 10% packet loss on both sides)] - [Performance Tests (IPv6, 10% packet loss on both sides)] - [InterOperability Tests (IPv4, msquic -> lkquic)] - [InterOperability Tests (IPv6, lkquic -> msquic)] - [Http/3 Tests (http3_test -> Public Websites)] - [Http/3 Tests (http3_test client -> http3_test server)] - [Session Resumption Tests] - [Sample Tests] - [ALPN and Preferred Address Tests] 2. Kernel Consumer Test via tlshd The testing can be run using `make check tests=tlshd` in libquic after tlshd service [2] is installed and configured. The tests include: - [Kernel Tests (kernel -> lkquic, Certificate, Sample)] - [Kernel Tests (lkquic -> kernel, Certificate, Sample)] - [Kernel Tests (kernel -> lkquic, PSK, Sample)] - [Kernel Tests (lkquic -> kernel, PSK, Sample)] - [Kernel Tests (kernel -> lkquic, Certificate, Session Resumption)] - [Kernel Tests (lkquic -> kernel, Certificate, Session Resumption)] 3. HTTP/3 Interoperability testing via curl Linux Kernel QUIC is being integrated for HTTP/3 in curl [4]. Below is an overview of connecting to various HTTP/3 servers using different QUIC implementations: # curl --http3-only --ipv4 https://cloudflare-quic.com/ # curl --http3-only --ipv4 https://facebook.com/ # curl --http3-only --ipv4 https://litespeedtech.com/ # curl --http3-only --ipv4 https://nghttp2.org:4433/ # curl --http3-only --ipv4 https://outlook.office.com/ # curl --http3-only --ipv4 https://www.google.com/ 4. Performance testing via iperf The performance testing was conducted using iperf [5] over a 100G physical NIC, evaluating various packet sizes and MTUs: - QUIC vs. kTLS: UNIT size:1024 size:4096 size:16384 size:65536 Gbits/sec QUIC | kTLS QUIC | kTLS QUIC | kTLS QUIC | kTLS ──────────────────────────────────────────────────────────────────── mtu:1500 1.67 | 2.16 3.04 | 5.04 3.49 | 7.84 3.83 | 7.95 ──────────────────────────────────────────────────────────────────── mtu:9000 2.17 | 2.41 5.47 | 6.19 6.45 | 8.66 7.48 | 8.90 - QUIC(disable_1rtt_encryption) vs. TCP: UNIT size:1024 size:4096 size:16384 size:65536 Gbits/sec QUIC | TCP QUIC | TCP QUIC | TCP QUIC | TCP ──────────────────────────────────────────────────────────────────── mtu:1500 2.17 | 2.49 3.59 | 8.36 6.09 | 15.1 6.92 | 16.2 ──────────────────────────────────────────────────────────────────── mtu:9000 2.47 | 2.54 7.66 | 7.97 14.7 | 20.3 19.1 | 31.3 The performance gap between QUIC and kTLS may be attributed to: - The absence of Generic Segmentation Offload (GSO) for QUIC. - An additional data copy on the transmission (TX) path. - Extra encryption required for header protection in QUIC. - A longer header length for the stream data in QUIC. NOTE: The QUIC module is currently labeled as "EXPERIMENTAL". [1] https://www.ietf.org/archive/id/draft-lxin-quic-socket-apis-00.html [2] https://github.com/oracle/ktls-utils [3] https://github.com/lxin/quic [4] https://github.com/moritzbuhl/curl (-b linux_curl) [5] https://github.com/lxin/iperf Xin Long (5): net: define IPPROTO_QUIC and SOL_QUIC constants for QUIC protocol net: include quic.h in include/uapi/linux for QUIC protocol net: implement QUIC protocol code in net/quic directory net: integrate QUIC build configuration into Kconfig and Makefile Documentation: introduce quic.rst to provide description of QUIC protocol Documentation/networking/quic.rst | 178 +++ include/linux/quic.h | 19 + include/linux/socket.h | 1 + include/uapi/linux/in.h | 2 + include/uapi/linux/quic.h | 192 +++ net/Kconfig | 1 + net/Makefile | 1 + net/quic/Kconfig | 34 + net/quic/Makefile | 19 + net/quic/cong.c | 630 +++++++++ net/quic/cong.h | 118 ++ net/quic/connid.c | 188 +++ net/quic/connid.h | 120 ++ net/quic/crypto.c | 996 +++++++++++++ net/quic/crypto.h | 153 ++ net/quic/frame.c | 1903 +++++++++++++++++++++++++ net/quic/frame.h | 198 +++ net/quic/hashtable.h | 145 ++ net/quic/input.c | 602 ++++++++ net/quic/input.h | 155 ++ net/quic/number.h | 314 +++++ net/quic/output.c | 748 ++++++++++ net/quic/output.h | 199 +++ net/quic/packet.c | 1523 ++++++++++++++++++++ net/quic/packet.h | 125 ++ net/quic/path.c | 422 ++++++ net/quic/path.h | 143 ++ net/quic/pnspace.c | 184 +++ net/quic/pnspace.h | 209 +++ net/quic/protocol.c | 950 +++++++++++++ net/quic/protocol.h | 71 + net/quic/socket.c | 2183 +++++++++++++++++++++++++++++ net/quic/socket.h | 267 ++++ net/quic/stream.c | 252 ++++ net/quic/stream.h | 150 ++ net/quic/test/sample_test.c | 615 ++++++++ net/quic/test/unit_test.c | 1190 ++++++++++++++++ net/quic/timer.c | 302 ++++ net/quic/timer.h | 43 + 39 files changed, 15545 insertions(+) create mode 100644 Documentation/networking/quic.rst create mode 100644 include/linux/quic.h create mode 100644 include/uapi/linux/quic.h create mode 100644 net/quic/Kconfig create mode 100644 net/quic/Makefile create mode 100644 net/quic/cong.c create mode 100644 net/quic/cong.h create mode 100644 net/quic/connid.c create mode 100644 net/quic/connid.h create mode 100644 net/quic/crypto.c create mode 100644 net/quic/crypto.h create mode 100644 net/quic/frame.c create mode 100644 net/quic/frame.h create mode 100644 net/quic/hashtable.h create mode 100644 net/quic/input.c create mode 100644 net/quic/input.h create mode 100644 net/quic/number.h create mode 100644 net/quic/output.c create mode 100644 net/quic/output.h create mode 100644 net/quic/packet.c create mode 100644 net/quic/packet.h create mode 100644 net/quic/path.c create mode 100644 net/quic/path.h create mode 100644 net/quic/pnspace.c create mode 100644 net/quic/pnspace.h create mode 100644 net/quic/protocol.c create mode 100644 net/quic/protocol.h create mode 100644 net/quic/socket.c create mode 100644 net/quic/socket.h create mode 100644 net/quic/stream.c create mode 100644 net/quic/stream.h create mode 100644 net/quic/test/sample_test.c create mode 100644 net/quic/test/unit_test.c create mode 100644 net/quic/timer.c create mode 100644 net/quic/timer.h