mchus/logpile
1
0
Fork 0
Code Issues 5 Pull Requests Actions Packages Projects Releases 9 Wiki Activity

docs: document firmware filter regression pattern in bible (ADL-019)

Browse Source 
Root cause analysis for device-bound firmware leaking into hardware.firmware
on Supermicro Redfish (SYS-A21GE-NBRT HGX B200):

- collectFirmwareInventory (6c19a58) had no coverage for Supermicro naming.
  isDeviceBoundFirmwareName checked "gpu " / "nic " (space-terminated) while
  Supermicro uses "GPU1 System Slot0" / "NIC1 System Slot0 ..." (digit suffix).

- 9c5512d added _fw_gpu_ / _fw_nvswitch_ / _inforom_gpu_ patterns to fix HGX,
  but checked DeviceName which contains "Software Inventory" (from Redfish Name),
  not the firmware Id. Dead code from day one.

09-testing.md: add firmware filter worked example and rule #4 — verify the
filter checks the field that the collector actually populates.

10-decisions.md: ADL-019 — isDeviceBoundFirmwareName must be extended per
vendor with a test case per vendor format before shipping.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Mikhail Chusavitin
2026-03-12 14:03:47 +03:00
parent 5815100e2f
commit 47bb0ee939
2 changed files with 47 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
bible-local/09-testing.md
View File

@@ -71,12 +71,34 @@ TestXxx_KeepsDistinct — two different items with same model → 2 res
TestXxx_<VendorThatMotivated> — the specific vendor/setup that triggered the code
```
### Worked example — firmware filter regression (2026-03-12)
`collectFirmwareInventory` was added in `6c19a58` without coverage for Supermicro naming.
`isDeviceBoundFirmwareName` had patterns for Dell-style names (`"GPU SomeDevice"`, `"NIC OnboardLAN"`)
but Supermicro Redfish uses `"GPU1 System Slot0"` and `"NIC1 System Slot0 ..."` — digit follows
immediately after the type prefix. 29 device-bound entries leaked into `hardware.firmware`.
`9c5512d` attempted to fix this with HGX ID patterns (`_fw_gpu_`, etc.) in the wrong field:
the filter checked `DeviceName` but `collectFirmwareInventory` populates it from `Name` first
(`"Software Inventory"` for all HGX per-component slots), not from the `Id` field that contains
the firmware ID like `"HGX_FW_GPU_SXM_1"`. The patterns were effectively dead code from day one.
**Required test matrix for any filter function:**
```
TestXxx_FiltersDeviceBound_Dell — Dell-style names that motivated the original code
TestXxx_FiltersDeviceBound_Supermicro — Supermicro names with digit suffix (GPU1/NIC1)
TestXxx_KeepsSystemLevel — BIOS, BMC, CPLD names must NOT be filtered
```
### Practical rule
When you write a new filter/dedup/classify function, ask:
1. Does my test cover the vendor that motivated this code?
2. Does my test cover a *different* vendor or naming convention where the function must NOT fire?
3. If I change the dedup key logic, do existing tests still exercise the old correct behavior?
4. When the filter checks a field on a model struct, does my test verify that the field is
actually populated by the collector? (Dead-code filter pattern: `9c5512d` `_fw_gpu_` check.)
If any answer is "no" — add the missing test before committing.