tailscale/tailscale

magicsock_unix_test fails when the machine already has higher buffer sizes

开放

#15,994 创建于 2025年5月16日

 (5 条评论) (0 个反应) (0 位负责人)Go (2,554 个派生)batch import
OS-linuxbuggood first issue

仓库指标

星标
 (31,565 个星标)
PR 合并指标
 (平均合并 5天 22小时) (30 天内合并 173 个 PR)

描述

What is the issue?

https://github.com/tailscale/tailscale/blob/9c52856af62074e51f125a90b4153fab61a13b00/wgengine/magicsock/magicsock_unix_test.go#L43-L47

The test applies the buffer size setting, and then checks that the buffer size goes up. If building Tailscale on a machine where the buffer size is already set higher than the Tailscale defaults, the buffer size doesn't go up - it goes down. The test then, obviously, fails.

Steps to reproduce

sysctl -w net.core.wmem_default=16777216, then build Tailscale

Are there any recent changes that introduced the issue?

No response

OS

Linux

OS version

NixOS unstable

Tailscale version

1.82.5

Other software

No response

Bug report

No response

贡献者指南