fix(tun): TUN routing failed on every connect, freezing under load

A user's log showed the same two failures on all 9 connects, 198 route errors
total:
  Added 0 bypass routes via 192.168.88.1 (if_index=11)
  Could not find ostp_tun index in routing table after 15s — traffic will NOT be captured

Two independent bugs in the Windows route layer:

1. The TUN interface index was looked up by matching FriendlyName == "ostp_tun"
   through GetAdaptersAddresses. WinTun does not set the FriendlyName to the
   adapter name, so the match never succeeded — every connect burned the full
   15s window and gave up. The tun crate hands the real index back directly via
   AbstractDevice::tun_index() (WinTun's own adapter index), which is instant and
   correct; the name lookup remains only as a fallback.

2. Every route add — the server-IP bypass and the TUN default route alike — went
   through the legacy CreateIpForwardEntry, which failed with error 160
   (ERROR_BAD_ARGUMENTS) on this machine for all of them. With the server-IP
   bypass never installed, the server's own packets were routed INTO the tunnel:
   a loop that stalls the link for seconds under load (the reported "VPN drops
   ~8s into a game" — the tunnel never actually disconnected, it froze; the log
   showed gap recovery skipping up to 402 frames with no packet loss on the
   wire). add_ipv4_route now shells to route.exe, which resolves the interface
   and validates the gateway itself and is already what the teardown path uses;
   its command form was verified to be accepted (fails only on elevation, not
   syntax). CREATE_NO_WINDOW keeps it from flashing a console per route.

Cannot be verified without the user's elevated TUN environment; the next log
will read "Added N bypass routes" and "Default route via TUN ... added" instead
of the failures.
This commit is contained in:
ospab 2026-08-18 23:13:51 +03:00
parent 44677c68e4
commit 321365efe3
2 changed files with 66 additions and 23 deletions

View File

@ -99,18 +99,34 @@ pub async fn create(opts: OstpTunOptions) -> Result<OstpTunInterface> {
let current_exe = std::env::current_exe()?.to_string_lossy().into_owned();
// A freshly created WinTun adapter can take several seconds to appear in
// GetAdaptersAddresses (it only shows up once it has an operational IPv4
// binding). The default route via the TUN is what actually captures
// traffic, so this lookup is critical — give it a generous window (~15s).
let mut tun_index = None;
for _ in 0..75 {
if let Some(idx) = windows_route::sys::get_interface_index("ostp_tun") {
tun_index = Some(idx);
break;
// Take the interface index straight from the adapter WinTun just created,
// via the tun crate. The old code looked it up by FriendlyName == "ostp_tun"
// through GetAdaptersAddresses — but WinTun does NOT set the FriendlyName to
// the adapter name, so that match never succeeded: on every single connect
// it spun the full 15s and then gave up with "traffic will NOT be captured",
// leaving the default route (and, above, the server-IP bypass) uninstalled.
// get_adapter_index() is instant and correct.
use tun::AbstractDevice;
let tun_index = match dev.tun_index() {
Ok(idx) if idx > 0 => Some(idx as u32),
Ok(idx) => {
tracing::error!("WinTun reported a non-positive interface index ({idx})");
None
}
tokio::time::sleep(std::time::Duration::from_millis(200)).await;
}
Err(e) => {
// Fall back to the old name lookup rather than fail outright.
tracing::warn!("Could not read TUN index from the adapter ({e}); falling back to name lookup");
let mut idx = None;
for _ in 0..75 {
if let Some(i) = windows_route::sys::get_interface_index("ostp_tun") {
idx = Some(i);
break;
}
tokio::time::sleep(std::time::Duration::from_millis(200)).await;
}
idx
}
};
if let Some(idx) = tun_index {
match windows_route::sys::add_ipv4_route(

View File

@ -17,7 +17,7 @@ pub mod sys {
use winapi::shared::minwindef::{DWORD, ULONG};
use winapi::shared::winerror::{ERROR_INSUFFICIENT_BUFFER, NO_ERROR};
use winapi::um::iphlpapi::{
CreateIpForwardEntry, DeleteIpForwardEntry, GetAdaptersAddresses, GetIpForwardTable,
DeleteIpForwardEntry, GetAdaptersAddresses, GetIpForwardTable,
};
use winapi::um::iptypes::{
GAA_FLAG_SKIP_ANYCAST, GAA_FLAG_SKIP_DNS_SERVER, GAA_FLAG_SKIP_MULTICAST, IP_ADAPTER_ADDRESSES,
@ -88,20 +88,47 @@ pub mod sys {
if_index: u32,
metric: u32,
) -> Result<(), String> {
let mut row: MIB_IPFORWARDROW = unsafe { mem::zeroed() };
row.dwForwardDest = ipv4_to_dword(dest);
row.dwForwardMask = ipv4_to_dword(mask);
row.dwForwardNextHop = ipv4_to_dword(nexthop);
row.dwForwardIfIndex = if_index;
row.ForwardType = if nexthop == Ipv4Addr::UNSPECIFIED || dest == nexthop { 3 } else { 4 };
row.ForwardProto = 3; // MIB_IPPROTO_NETMGMT
row.dwForwardMetric1 = metric;
// Installed through route.exe rather than CreateIpForwardEntry.
//
// The legacy CreateIpForwardEntry API was failing here with error 160
// (ERROR_BAD_ARGUMENTS) on every single route — server-IP bypass and TUN
// default route alike — which left the server IP routed INTO the tunnel
// (a loop that froze the link for seconds under load) and the default
// route uninstalled. route.exe resolves the interface and validates the
// gateway itself, and is what the teardown path already uses, so it
// succeeds where the hand-built MIB_IPFORWARDROW did not.
use std::os::windows::process::CommandExt;
use std::process::Command;
const CREATE_NO_WINDOW: u32 = 0x0800_0000;
let ret = unsafe { CreateIpForwardEntry(&mut row) };
if ret == NO_ERROR {
// route add <dest> mask <mask> <gateway> metric <m> if <ifindex>
let out = Command::new("route")
.args([
"add",
&dest.to_string(),
"mask",
&mask.to_string(),
&nexthop.to_string(),
"metric",
&metric.to_string(),
"if",
&if_index.to_string(),
])
.creation_flags(CREATE_NO_WINDOW)
.output()
.map_err(|e| format!("could not run route.exe: {e}"))?;
if out.status.success() {
Ok(())
} else {
Err(format!("CreateIpForwardEntry failed: {}", ret))
// route.exe prints its diagnostics to stdout, not stderr.
let msg = String::from_utf8_lossy(&out.stdout);
let msg = msg.trim();
Err(format!(
"route add failed (exit {:?}): {}",
out.status.code(),
if msg.is_empty() { "no output" } else { msg }
))
}
}