Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

examine: lspci cannot name a GPU that pci.ids does not know, though the device id is on the line

Open Beginner friendly
#507 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
2/5
Estimated time
1-3 hours
Newbie friendliness
68/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Active
Tech stack
linux, rust

Research direction

Start at extract_lspci_device_id and gfx_target_from_amd_pci_device_id, following the Linux lspci path and the existing Windows device-id behavior. Verify the hardware ids from lspci -nn where possible, then add confirmed mappings for 1586, 1636, 1638, 164c, and 15e7; done means rocm examine --json resolves the corresponding gfx_target.

Written by the indexing model from the issue text.

Description

bug

Current behavior

Every lspci -nn line carries the vendor:device id in brackets, e.g.

c5:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Raphael [1002:164e] (rev cb)

When the host's pci.ids database has no entry for the device, lspci prints the bare word Device in place of the name. Classification then has nothing to work with and the GPU resolves to no target — even though [1002:xxxx] is right there on the same line and the crate already has a device-id decoder.

Expected behavior

The device id is preferred over the marketing name, as it is on the Windows probe path, which reads the id out of the PNP id.

Steps to reproduce

On a host whose pci.ids predates the GPU — the reported case is Strix Halo, see #449 — run rocm examine --json and observe an empty gfx_target.

Status

The reader half is now in place: extract_lspci_device_id pulls [1002:xxxx] from the line (AMD vendor only, so the [0300] class tag cannot be mistaken for a device id) and is preferred over the name. What is missing is table coverage: gfx_target_from_amd_pci_device_id has no entry for

  • 1586 — Strix Halo (the device in #449)
  • 1636, 1638, 164c, 15e7 — the gfx90c parts (Renoir, Cezanne, Lucienne, Barcelo)

These were deliberately not added without hardware to confirm them: guessing numeric ids into a public hardware table is worse than a documented gap. Anyone with the part can verify the id from lspci -nn and add the row.

Environment

  • Linux. The equivalent Windows path already reads the device id.

Additional context

Related to #449, which reports the user-visible symptom on Strix Halo. This issue covers the mechanism and what remains to close it.

Dominant language
Rust
Stars
45
Forks
11
Avg merge
4d 9h
Merged PRs (30d)
107

Getting set up

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.

More from ROCm/rocm-cli

All issues in ROCm/rocm-cli

Similar issues

More Rust issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.