| Message ID | 20240425031340.46946-1-kerneljasonxing@gmail.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | Implement reset reason mechanism to detect | expand |
On Thu, Apr 25, 2024 at 5:13 AM Jason Xing <kerneljasonxing@gmail.com> wrote: > > From: Jason Xing <kernelxing@tencent.com> > > In production, there are so many cases about why the RST skb is sent but > we don't have a very convenient/fast method to detect the exact underlying > reasons. > > RST is implemented in two kinds: passive kind (like tcp_v4_send_reset()) > and active kind (like tcp_send_active_reset()). The former can be traced > carefully 1) in TCP, with the help of drop reasons, which is based on > Eric's idea[1], 2) in MPTCP, with the help of reset options defined in > RFC 8684. The latter is relatively independent, which should be > implemented on our own, such as active reset reasons which can not be > replace by skb drop reason or something like this. > > In this series, I focus on the fundamental implement mostly about how > the rstreason mechanism works and give the detailed passive part as an > example, not including the active reset part. In future, we can go > further and refine those NOT_SPECIFIED reasons. > > Here are some examples when tracing: > <idle>-0 [002] ..s1. 1830.262425: tcp_send_reset: skbaddr=x > skaddr=x src=x dest=x state=x reason=NOT_SPECIFIED > <idle>-0 [002] ..s1. 1830.262425: tcp_send_reset: skbaddr=x > skaddr=x src=x dest=x state=x reason=NO_SOCKET > > [1] > Link: https://lore.kernel.org/all/CANn89iJw8x-LqgsWOeJQQvgVg6DnL5aBRLi10QN2WBdr+X4k=w@mail.gmail.com/ Reviewed-by: Eric Dumazet <edumazet@google.com> Thanks !
Hello: This series was applied to netdev/net-next.git (main) by Paolo Abeni <pabeni@redhat.com>: On Thu, 25 Apr 2024 11:13:33 +0800 you wrote: > From: Jason Xing <kernelxing@tencent.com> > > In production, there are so many cases about why the RST skb is sent but > we don't have a very convenient/fast method to detect the exact underlying > reasons. > > RST is implemented in two kinds: passive kind (like tcp_v4_send_reset()) > and active kind (like tcp_send_active_reset()). The former can be traced > carefully 1) in TCP, with the help of drop reasons, which is based on > Eric's idea[1], 2) in MPTCP, with the help of reset options defined in > RFC 8684. The latter is relatively independent, which should be > implemented on our own, such as active reset reasons which can not be > replace by skb drop reason or something like this. > > [...] Here is the summary with links: - [net-next,v9,1/7] net: introduce rstreason to detect why the RST is sent https://git.kernel.org/netdev/net-next/c/5cb2cb3cb20c - [net-next,v9,2/7] rstreason: prepare for passive reset https://git.kernel.org/netdev/net-next/c/6be49deaa095 - [net-next,v9,3/7] rstreason: prepare for active reset https://git.kernel.org/netdev/net-next/c/5691276b39da - [net-next,v9,4/7] tcp: support rstreason for passive reset https://git.kernel.org/netdev/net-next/c/120391ef9ca8 - [net-next,v9,5/7] mptcp: support rstreason for passive reset https://git.kernel.org/netdev/net-next/c/3e140491dd80 - [net-next,v9,6/7] mptcp: introducing a helper into active reset logic https://git.kernel.org/netdev/net-next/c/215d40248bde - [net-next,v9,7/7] rstreason: make it work in trace world https://git.kernel.org/netdev/net-next/c/b533fb9cf4f7 You are awesome, thank you!
From: Jason Xing <kernelxing@tencent.com> In production, there are so many cases about why the RST skb is sent but we don't have a very convenient/fast method to detect the exact underlying reasons. RST is implemented in two kinds: passive kind (like tcp_v4_send_reset()) and active kind (like tcp_send_active_reset()). The former can be traced carefully 1) in TCP, with the help of drop reasons, which is based on Eric's idea[1], 2) in MPTCP, with the help of reset options defined in RFC 8684. The latter is relatively independent, which should be implemented on our own, such as active reset reasons which can not be replace by skb drop reason or something like this. In this series, I focus on the fundamental implement mostly about how the rstreason mechanism works and give the detailed passive part as an example, not including the active reset part. In future, we can go further and refine those NOT_SPECIFIED reasons. Here are some examples when tracing: <idle>-0 [002] ..s1. 1830.262425: tcp_send_reset: skbaddr=x skaddr=x src=x dest=x state=x reason=NOT_SPECIFIED <idle>-0 [002] ..s1. 1830.262425: tcp_send_reset: skbaddr=x skaddr=x src=x dest=x state=x reason=NO_SOCKET [1] Link: https://lore.kernel.org/all/CANn89iJw8x-LqgsWOeJQQvgVg6DnL5aBRLi10QN2WBdr+X4k=w@mail.gmail.com/ v9 Link: https://lore.kernel.org/all/20240423072137.65168-1-kerneljasonxing@gmail.com/ 1. address nit problem (Matt) 2. add acked-by and reviewed-by tags (Matt) v8 Link: https://lore.kernel.org/all/20240422030109.12891-1-kerneljasonxing@gmail.com/ 1. put sk reset reasons into more natural order (Matt) 2. adjust those helper position (Matt) 3. rename two convert function (Matt) 4. make the kdoc format correct (Simon) v7 Link: https://lore.kernel.org/all/20240417085143.69578-1-kerneljasonxing@gmail.com/ 1. get rid of enum casts which could bring potential issues (Eric) 2. use switch-case method to map between reset reason in MPTCP and sk reset reason (Steven) 3. use switch-case method to map between skb drop reason and sk reset reason v6 1. add back casts, or else they are treated as error. v5 Link: https://lore.kernel.org/all/20240411115630.38420-1-kerneljasonxing@gmail.com/ 1. address format issue (like reverse xmas tree) (Eric, Paolo) 2. remove unnecessary casts. (Eric) 3. introduce a helper used in mptcp active reset. See patch 6. (Paolo) v4 Link: https://lore.kernel.org/all/20240409100934.37725-1-kerneljasonxing@gmail.com/ 1. passing 'enum sk_rst_reason' for readability when tracing (Antoine) v3 Link: https://lore.kernel.org/all/20240404072047.11490-1-kerneljasonxing@gmail.com/ 1. rebase (mptcp part) and address what Mat suggested. v2 Link: https://lore.kernel.org/all/20240403185033.47ebc6a9@kernel.org/ 1. rebase against the latest net-next tree Jason Xing (7): net: introduce rstreason to detect why the RST is sent rstreason: prepare for passive reset rstreason: prepare for active reset tcp: support rstreason for passive reset mptcp: support rstreason for passive reset mptcp: introducing a helper into active reset logic rstreason: make it work in trace world include/net/request_sock.h | 4 +- include/net/rstreason.h | 121 +++++++++++++++++++++++++++++++++++++ include/net/tcp.h | 3 +- include/trace/events/tcp.h | 26 ++++++-- net/dccp/ipv4.c | 10 +-- net/dccp/ipv6.c | 10 +-- net/dccp/minisocks.c | 3 +- net/ipv4/tcp.c | 15 +++-- net/ipv4/tcp_ipv4.c | 17 ++++-- net/ipv4/tcp_minisocks.c | 3 +- net/ipv4/tcp_output.c | 5 +- net/ipv4/tcp_timer.c | 9 ++- net/ipv6/tcp_ipv6.c | 20 +++--- net/mptcp/protocol.c | 2 +- net/mptcp/protocol.h | 38 ++++++++++++ net/mptcp/subflow.c | 27 ++++++--- 16 files changed, 266 insertions(+), 47 deletions(-) create mode 100644 include/net/rstreason.h