SagerNet / SagerNet/sing-snell

snellv4: reuse 场景下 vectorised 写路径 panic buffer overflow: capacity N, start 0, need 23(record writer 叠层时缺 headroom 处理;snellv6 存在同类路径)

Open
#4 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Go
Stars
21
Forks
7
PR merge metrics
No merged PRs in 30d

Description

环境

  • sing-snell a4e97ee(v0.0.0-20260710094516-a4e97ee24beb),经 sing-box snell 出站使用
  • linux/amd64,Go 1.26
  • 出站配置形如:{"type":"snell","version":4,"reuse":true,"tcp_fast_open":true,...}

现象

真实流量下每几分钟必现 panic 进程崩溃。多次崩溃 capacity 各异(78 / 791 / 1067 / 4302 / 5054 / 9986),start 恒为 0、need 恒为 23:

panic: buffer overflow: capacity 791,start 0, need 23

goroutine 2193 [running]:
github.com/sagernet/sing/common/buf.(*Buffer).ExtendHeader(0x2dcc609fea80, 0x17)
sing@v0.8.12-0.20260702081104-2ded2af32d3d/common/buf/buffer.go:116
github.com/sagernet/sing-snell/snellv4.(*writer).makeBufferRecordLocked(0x2dcc5f9a6a20, 0x2dcc609fea80, 0x0)
sing-snell@v0.0.0-20260710094516-a4e97ee24beb/snellv4/record.go:417
github.com/sagernet/sing-snell/snellv4.(*vectorisedWriter).WriteVectorised(0x2dcc5f7a9b90, {0x2dcc5ec4d688, 0x4a, 0x8f})
snellv4/record.go:657
github.com/sagernet/sing-snell/snellv4.(*vectorisedWriter).WriteVectorised(0x2dcc5f7a9c50, {0x2dcc609fe740, 0x7, 0x7})
snellv4/record.go:684
github.com/sagernet/sing/common/bufio.copyWaitVectorisedWithPool(...)
common/bufio/copy_direct.go:87
github.com/sagernet/sing/common/bufio.copyWaitWithPool(...)
common/bufio/copy_direct.go:65
github.com/sagernet/sing/common/bufio.copyExtended(...)
common/bufio/copy.go:128
github.com/sagernet/sing/common/bufio.CopyWithCounters(...)
common/bufio/copy.go:99
github.com/sagernet/sing/common/bufio.CopyWithIncreateBuffer(...)
common/bufio/copy.go:58
github.com/sagernet/sing-box/route.(*ConnectionManager).connectionCopy(...)
route/conn.go:274
created by github.com/sagernet/sing-box/route.(*ConnectionManager).NewConnection
route/conn.go:152

注意栈中有两个 snellv4.vectorisedWriter 实例::684 帧(外层)收到 pool 的 7 个 buffer、做完 records 后正在
w.upstream.WriteVectorised(flushRecords);:657 帧(内层)收到这 74 个 record,在其中一个上调 makeBufferRecordLocked 时崩溃。

任何形成两层 record writer 叠置的配置均可触发:reuse: true,或 snell 出站 detour 另一 snell 出站(实测两者均崩);reuse: false
且无叠层后不再出现。

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with snellv4/record.go at makeBufferRecordLocked and the nested vectorisedWriter.WriteVectorised calls around lines 417, 657, and 684. Read sing/common/buf.Buffer.ExtendHeader and reproduce with reuse enabled or two layered Snell outbounds. Done means layered record writers no longer panic with a buffer overflow while reuse traffic is running.

Written by the indexing model from the issue text.

Assessment

Tech stack
go
Domain
networking
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
47/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.