Incorrect calculation of recv throughput in tcptop.py

描述

int tcp_recvmsg(...) {
     ...
     int copied = 0;
     ...
     do {
         ...
         tcp_cleanup_rbuf(sk, copied);
         ...
         copied += used;
         ...
     } while (len > 0);
     
     tcp_cleanup_rbuf(sk, copied);
     return copied;
     ...
}

貢獻者指南

技術棧

pythonc

領域

performanceobservability

議題類型

bug

難度面向新貢獻者的預計實作難度，1 表示很小改動，5 表示專家級工作。

4

預計時間有經驗貢獻者完成調查、實作、測試並準備 pull request 的粗略時間範圍。

1-3 hours

活動狀態議題目前的可參與程度：新鮮、活躍、陳舊、阻塞或等待維護者輸入。

active

清晰度議題是否清楚說明預期改動、驗收標準和下一步。

clear

前置要求

TCP kernel internalsBPF tracingPython scriptingC code reading

新手友善度1-100 的估計分數，表示該議題對首次貢獻者的友善程度。

35

研究方向

Investigate the tcptop.py tool (likely in tools/tcptop.py) to understand how it hooks into tcp cleanup rbuf. The kernel code shows tcp cleanup rbuf is called multiple times with cumulative copied, leading to overcounting. The fix may involve tracking the last copied value or only using the final call. Check the BPF program attached to tcp cleanup rbuf and adjust the throughput calculation to avoid double counting. See comments in the issue for potential solutions.