GUID collisions (in FV names)

[orangecms]

(Firmware Volume) GUIDs are reappearing, contrary to their meaning. 💣

Impact

This causes an issue when selecting a hoisted FV to jump to when clicking the corresponding button in the file holding the FV image. We cannot use the GUID as a sufficient means to determine which FV to jump to.

Handling / TODO

Should we care at all? Should we just assume that collisions never happen and only full images are used? Can a similar situation occur when capsule updates are staged?

Details

See https://github.com/orangecms/utk-web/blob/uefi/src/fixtures/E7C52AMS.360.utk.json. Those FVs have the same GUID here, but not in the fixture from uefi-firmware-parser. It might be that I missed printing them out since uefi-firmware-parser JSON support is work in progress. Its fixture is at https://github.com/orangecms/utk-web/blob/uefi/src/fixtures/E7C52AMS.360.ufp.json.

ag 4F1C52D3 src/fixtures/E7C52AMS.360.* -B 14
src/fixtures/E7C52AMS.360.utk.json
323-        "Length": 7278592,
324-        "Signature": 1213613663,
325-        "Attributes": 327423,
326-        "HeaderLen": 72,
327-        "Checksum": 53007,
328-        "ExtHeaderOffset": 96,
329-        "Revision": 2,
330-        "Blocks": [
331-          {
332-            "Count": 1777,
333-            "Size": 4096
334-          }
335-        ],
336-        "FVName": {
337:          "GUID": "4F1C52D3-D824-4D2A-A2F0-EC40C23C5916"
--
33210-        "Length": 10153984,
33211-        "Signature": 1213613663,
33212-        "Attributes": 327423,
33213-        "HeaderLen": 72,
33214-        "Checksum": 60454,
33215-        "ExtHeaderOffset": 96,
33216-        "Revision": 2,
33217-        "Blocks": [
33218-          {
33219-            "Count": 2479,
33220-            "Size": 4096
33221-          }
33222-        ],
33223-        "FVName": {
33224:          "GUID": "4F1C52D3-D824-4D2A-A2F0-EC40C23C5916"

Cause

How does this happen?

The exact reason is not clear. However, we do know that this is a file extracted from an upgrade, so several assumptions do not hold:

• the image is not necessarily a full image, yet it contains firmware volumes etc
• the image is not necessarily a single image, but can be multiple images in one, e.g. for board variants
• the uprade is not necessarily a one-shot thing, but can be cumulative