graymon
graymon
This is because pwru fetches pid in kernel space while converts pid to process name in user space, so if the process sending skb is terminated before pwru starts the...
I was thinking using a user-space map[skb]pname to cache pname conversion, but that actually doesn't make sense because an skb could be transferred between processes (e.g. from a client process...
@Asphaltt I was thinking of focusing on the the sk which will bind / has been bound to an skb. It's more like pwru can retrieve the skb from sk,...
Considering that Cilium heavily relies on skb->cb, I can't tolerate this issue any longer! Do you think it's a good idea to fetch and output skb->cb when `--output-meta` is set?...
> We'd need to fix bpf_snprintf_btf() to retrieve the content correctly though, if I understand correctly? True, although it would be long way to go.
Thanks! I found it may be a technical obstacle for us to trace a **frame**'s lifetime, it seems no way to link an xdp_buff to an skb: the "skb addr"...
试试这个： https://github.com/daeuniverse/dae/pull/512
嗨, 我们怀疑可能是 MTU 丢包重传导致 100ms sniff 超时，我提供两种方法来排查。 方法一：抓包 如果你系统流量不大的话可以直接 `tcpdump -ni dae0 -w dae0.pcap` ，在抓包的同时触发一次 sniff 失败的请求。 如果系统流量较大，可以先 dig 看看 play.googleapis.com 能解析出哪些 ipv4，比如能解析出 142.251.175.95 和 142.251.12.95 的话就可以用 ip 过滤： `tcpdump...
@jchen290 肉眼没发现重传，你延长 sniff timeout 没用的话说明不是我们之前推测的原因（握手之后的第一个payload被丢包了）
@sumire88 Not really, this issue if for record and awarenes in case other users hit it.