sc07kvm
sc07kvm
I can't find that script and reproduce the situation =(
Of course, on newer kernels(>=5.11) this happens via the tcp_recvmsg_locked function, which is inlined: ```c static int tcp_recvmsg_locked() { ... do { ... tcp_cleanup_rbuf(...); ... } while (...); tcp_cleanup_rbuf(...); ......
tcpdrop.py hooks **tcp_cleanup_rbuf** to count the amount of data received based on the **"copied"** argument: ```c int kprobe__tcp_cleanup_rbuf(..., int copied) { ... ipv4_recv_bytes.increment(ipv4_key, copied); .... } ``` For example, if...
I have no idea how to fix this at all
> * Add two bpf progs to trace tcp_recvmsg() entry/exit. you can't do that. kretprobe has a limit on the number of functions that can be executed simultaneously ([maxactive](https://docs.kernel.org/trace/kprobes.html#how-does-a-return-probe-work) parameter)....
Using fentry/fexit will greatly reduce the scope of supported OS. Kernel >= v5.5 + CONFIG_DEBUG_INFO_BTF flag