hev-socks5-tunnel stops working after a while

[patterniha]

when using hev-socks5-tunnel, the app stop working after a while (no new connections can be established, but existing connections remain active) and you have to restart service so that everything will be okay.

This bug doesn't exist when you use badvpn or when you are in proxy mode.

changing Hev-tun-read/write-timeout value doesn't help either.

apart from myself, many other users have reported this issue.

[fuilloi]

v2rayNG版本号。使用配置。复现步骤。hevtun可以开启debug日志模式。

[patterniha]

v1.10.28, config is not important.

I've encountered it a few times before, and someone reported it today, so I decided to report this.
But now I couldn't reproduce the bug again.

Anyway, I'm not sure if the problem is related to hev-socks5-tunnel, but whenever the app stop working, it was in hev-socks5-tunnel mode.

If I encounter this problem again and I'm sure I'll reopen it again.

[t-e-s-tweb]

Yeah. Same. I had similar issue when there are ads on websites and a lot of requests are running. Can't exactly reproduce it but switching to older tunnel removes the problem.

[patterniha]

Yeah. Same. I had similar issue when there are ads on websites and a lot of requests are running. Can't exactly reproduce it but switching to older tunnel removes the problem.

right, I also have to restart service occasionally to get it to work properly, something is definitely wrong with hev-socks5-tunnel.

[2dust]

问题是存在的，但是没有提供一点有用的信息的 issue 是解决不了问题的。
请收集足够定位问题的信息，我们可以反馈给 hev tun 开发者

[patterniha]

The problem exists, but an issue without providing any useful information won't solve it. Please gather enough information to pinpoint the problem so we can forward it to the HEV TUN developers.

I didn't find any useful information in the log.
My guess is that when we have a lot of connections, we have memory leak because after it crashes, if you don't open any websites for a while, it will fix itself, although I didn't test it very thoroughly and more investigation is needed.

[cchkx]

My guess is that when we have a lot of connections, we have memory leak because after it crashes, if you don't open any websites for a while, it will fix itself

IIUC, hev-socks5-tunnel uses more file descriptors than badvpn-tun2socks. That suggests the issue is likely caused by reaching the RLIMIT_NOFILE limit. This would also explain why everything returns to normal after a period of inactivity, once idle sessions time out, the corresponding file descriptors are released and new sessions can be established.

Since the default RLIMIT_NOFILE value may vary across Android versions, it might be worth increasing it within the app. Another option is to use hev-socks5-tunnel's max-session-count to cap the number of active sessions and proactively close the least active ones to avoid hitting the file descriptor limit. We could also try running the tun2socks service in a separate process by configuring it in AndroidManifest.xml.

[cchkx]

@heiher Could you take a look? Thanks!

[heiher]

A single session in hev-socks5-tunnel does not use more file descriptors than badvpn-tun2socks. However, frequent DNS lookups may crate a large number of UDP sessions. If these sessions have relatively long timeouts, they can indeed result in high file descriptor usage.

We can try lowering the RLIMIT_NOFILE limit to see whether the issue becomes easier to reproduce. If the failure mode remains the same, that would strongly suggest this is the root cause.

[languang666]

Me too

[Saeed1388]

Do not show ip , because