Hacktoberfest 2026:メンテナが10月に向けて印を付けた、オープンで初心者向けの issue。 Hacktoberfest の issue を見る

elfuse segfaults before its first vCPU exists under a 1.5 TiB RLIMIT_AS

オープン
#459 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 3 日以内に返信

まだ誰も着手していません。

評価

難易度
4/5
見積もり時間
3〜5日
初心者へのやさしさ
48/100
issue の種類
バグ
明瞭さ
おおむね明確
活発さ
活発
技術スタック
c, macos

調査の方向性

Start with the startup trace around src/core/bootstrap.c:710 and inspect the code between build_linux_stack and hv_vcpu_create; the issue leaves the faulting instruction and whether Hypervisor.framework is responsible unresolved. Reproduce on macOS 26.4.1 with ulimit -v at 1.5 TiB, then compare with 6 TiB and macOS 27. Done means the fault is identified and elfuse either boots or exits with an error instead of segfaulting.

索引モデルが issue の本文から書いたものです。

説明

bug

Background

#448's test-rlimit-seed lane starts elfuse under a host ulimit -v, and at 1.5 TiB it exited 139 on the macOS 26.4.1 CI runner. Draft PR #457 found two separate crashes behind that; #458 has the full background and the table of runs. This issue is the one at 1.5 TiB, where guest memory is set up but boot never reaches the guest.

Intention

Under a host address-space limit that leaves too little room for a vCPU, elfuse should fail with an error and a non-zero exit, or boot, not segfault. The first step is to find what faults.

Evidence

Same #457 job and runner as #458: macOS 26.4.1 (25E253), Mac16,10 (M4), 24 GiB, ELFUSE_STARTUP_TRACE=1. Under ulimit -v 1.5 TiB, busybox true, --verbose busybox cat /proc/self/limits and test-proc-limits all exited 139, and so did the original lane case in 3 of 3 repeats and when run outside a command substitution. elfuse --version exited 0. At 6 TiB every run exited 0.

The slab fits and boot gets through build_linux_stack; the trace line for hv_vcpu_create (src/core/bootstrap.c:710) never appears:

startup primary_mmap                    0.700 ms
startup cow_shm_upgrade                 1.361 ms
startup hv_vm_map                       2.560 ms
startup guest_init                      4.695 ms
...
startup vdso_build                      0.002 ms
startup build_linux_stack               0.011 ms
Segmentation fault: 11

At 6 TiB the same trace continues with startup hv_vcpu_create 0.437 ms and the guest runs.

hv_vcpu_create is wrapped in HV_CHECK, which prints a crash report when the call returns an error, and none was printed.

Not confirmed:

  • Which instruction faults. The runner left no crash report under ~/Library/Logs/DiagnosticReports for these processes.
  • Whether the fault is inside Hypervisor.framework. The missing HV_CHECK report suggests hv_vcpu_create faults rather than returns an error when the limit leaves it no room, but the code between build_linux_stack and that call is not ruled out.
  • Why this needs so much address space on macOS 26. On macOS 27 (M-series) the same build boots under 1.5 TiB.

Refs #448, #457, #458

主要言語
C
スター
275
フォーク
29
平均マージ
2日 22時間
マージ済み PR(30日)
18

環境構築

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

sysprog21/elfuse のほかの issue

sysprog21/elfuse の issue をすべて見る

似ている issue

C の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。