An AI agent for expert analysis of Linux kernel OOPS crash reports on x86. The skill gives an agent a structured, multi-step workflow for extracting structured data from a crash report and reasoning...
概览
An AI agent for expert analysis of Linux kernel OOPS crash reports on x86. The skill gives an agent a structured, multi-step workflow for extracting structured data from a crash report and reasoning toward a root cause hypothesis. Real-world crash reports analysed end-to-end by the skill: Once installed, you can talk to the agent naturally. A few examples: A customer reported a kernel oops — I saved the report in crash.txt. Please analyse it and provide recommendations. My local system had a kernel crash. Can you pull the crash out of dmesg and provide a thorough analysis? Ubuntu bug 2148595 has a kernel crash. Can you fetch it and analyse the oops? (see result) Please analyse the oops in kernel.org bugzilla nr 221376. Debian bug 939277 contains a kernel oops. Can you dig into the root cause?
README
以下为原文 README 节选，完整内容请查看原文。
Linux Kernel OOPS Analysis — AI Agent Skill
An AI agent skill for expert analysis of Linux kernel OOPS crash reports on x86. The skill gives an agent a structured, multi-step workflow for extracting structured data from a crash report and reasoning toward a root cause hypothesis.
Examples
Real-world crash reports analysed end-to-end by the skill:
| Oops report | Analysis | Description |
|---|---|---|
| fedora-oops.txt | example-1.md | Fedora 43, ZhiWei NA08H tablet — bmc150_accel_core NULL pointer dereference on interrupt-less hardware; upstream fix in v6.18 |
| Launchpad #2134472 | example-2.md | Ubuntu 24.04 Noble, GCP — l3mdev_fib_table_rcu NULL dereference during VRF teardown; race with ARP neighbour timer in IRQ context |
| lkml msgid aYN3JC_Kdgw5G2Ik@861G6M3 | example-3.md | Cloudflare Linux 6.18.7 — VM_BUG_ON_FOLIO in filemap_fault; THP folio split race placing sub-folios at wrong XArray indices; fix in v7.0-rc1, not yet in v6.18 stable |
| bugzilla.kernel.org #221376 | example-4.md | Custom 6.18.22 kernel, AMD RX 9070 XT (RDNA4/GFX12) — DRM_MM_BUG_ON in drm_mm_init; RDNA4 removed GDS/GWS/OA resources but amdgpu_ttm_init unconditionally passes zero-size ranges; no upstream fix found as of v6.19 |
| lkml CAPHJ_VJeBAL_fk+… | example-5.md · patch email | Mainline 7.0.0-08391-g1d51b370a0f8, syzkaller ext4 image — BUG_ON(pos+len > i_inline_size) in ext4_write_inline_data; corrupted image bypasses inline-size guard; fix replaces BUG_ON with ext4_error_inode() |
Example prompts
Once installed, you can talk to the agent naturally. A few examples:
From a saved file
A customer reported a kernel oops — I saved the report in
crash.txt. Please analyse it and provide recommendations.
From the live system
My local system had a kernel crash. Can you pull the crash out of
dmesgand provide a thorough analysis?
From a Launchpad bug (Ubuntu)
Ubuntu bug 2148595 has a kernel crash. Can you fetch it and analyse the oops?
From the kernel Bugzilla (see result)
Please analyse the oops in kernel.org bugzilla nr 221376.
From the Debian bug tracker
Debian bug 939277 contains a kernel oops. Can you dig into the root cause?
From Red Hat / Fedora Bugzilla
Fedora bug 1718414 on bugzilla.redhat.com shows a crash — what’s wrong?
From an lkml / lore.kernel.org email (see result)
推荐工具
换一个关键词，或者移除筛选条件。
安装
npx skillfish add intel/linux-kernel-oops