Remove internal/external disk type from PhysicalDisk view

[zephraph]

This PR follows up on oxidecomputer/console#1628. I started #3482 to clarify the field but it was decided there it's best just to take it out.

[zephraph]

Changed this to a U2 b/c the M2 is filtered out in the results

[smklein]

Personally I find this a little confusing, an internal function called sled_list_physical_disks doesn't indicate that it would filter out M.2s.

If we're keeping that name: we should continue inserting this M.2 in the test, and keep returning it.
Otherwise, we should rename sled_list_physical_disks to sled_list_external_disks or something.

[zephraph]

I'm fine w/ changing the name if that helps. We can't leave the M.2s in because sled_list_physical_disk pipes up to the app layer sled_list_physical_disk which maps to /v1/system/hardware/sleds/:sled/disks.

[rmustacc]

Why wouldn't we just list both the U.2s and M.2s? The original reason we had confusion was that I think @ericaasen saw the wrong types here. This endpoint feels like it should list everything. We talked in #3482 about terminology challenges, but I don't see why wouldn't list all disks here.

[zephraph]

This is the part that verifies the filtering works

[david-crespo]

Looks good. Only part that worries me is the potential for confusion based on the expectation that we would be listing everything in the DB rather than filtering for U2s. Short of changing the name of the struct, maybe a comment on the list endpoint would help? Or maybe changing the name of the Nexus function? I'm kind of flailing, but I feel like some name somewhere should give a hint that that filter is present in the query.

omicron/nexus/src/external_api/http_entrypoints.rs

Lines 4248 to 4254 in 9eac70a

	/// List physical disks
	#[endpoint {
	method = GET,
	path = "/v1/system/hardware/disks",
	tags = ["system/hardware"],
	}]
	async fn physical_disk_list(

[zephraph]

Chatted offline with @rmustacc about the filtering. I'd slightly misunderstood the original issue so I've removed all the filtering. U.2s and M.2s will be listed but the type field is still out.

[david-crespo]

Solves my problem!

[smklein]

Chatted offline with @rmustacc about the filtering. I'd slightly misunderstood the original issue so I've removed all the filtering. U.2s and M.2s will be listed but the type field is still out.

I'm unclear on this -- how is it less confusing to expose both U.2s and M.2s, but make them totally indistinguishable from one another?

[smklein]

Follow-up from some out-of-band discussions:

• Exposing the info "external" vs "internal" is too opinionated, and not a sufficient replacement of true topology information
• However, exposing the form factor of the device -- u.2 vs m.2 -- is a "fact". We'd like to expose more topology info when we can, but providing this info will be helpful when operating on disks in the short-term, when we have a single gimlet layout.

Therefore: We should probably still report whether or not disks are u.2s or m.2s, and we can provide more topology information later too.