[shimV2] refactor scsi device controller to V2 template

[rawahars]

This PR is mostly a structural refactor + type renames + logging.

Summary of what was done (from code changes)

A) Reorganized the SCSI code into “template-style” modules

Moved definitions out of the main implementation files into dedicated files:

• internal/controller/device/scsi/

  • Added doc.go (package docs)
  • Added types.go (interfaces + numLUNsPerController + reservation struct)
  • Deleted scsi.go (old minimal package comment file)

• internal/controller/device/scsi/disk/

  • Added types.go (disk Type, disk Config, VM/guest interfaces)
  • Added state.go (disk lifecycle state enum + String())
  • Added doc.go (disk lifecycle documentation)
  • Added disk_lcow.go and disk_wcow.go (currently only //go:build windows + package disk; effectively placeholders)

• internal/controller/device/scsi/mount/

  • Added types.go (mount Config + guest interfaces + Equals)
  • Added state.go (mount lifecycle state enum + String())
  • Added doc.go (mount lifecycle documentation)

Net effect: same concepts, but split into consistent “types/state/doc” layout.

B) Renamed public types

• disk.DiskConfig → disk.Config
• disk.DiskType* → disk.Type*
• disk.DiskState* → disk.State*
• mount.MountConfig → mount.Config
• mount.MountState* → mount.State*

Also Controller’s reservation ID type changed:

• Controller.Reserve(...) returns guid.GUID instead of uuid.UUID
• Controller.MapToGuest/UnmapFromGuest(...) now take guid.GUID instead of uuid.UUID
• Internally: reservations map[guid.GUID]*reservation

C) Added structured logging + new logfields

Changes are sprinkled through Controller, disk.Disk, and mount.Mount:

• internal/logfields/fields.go adds:

  • Controller, LUN, DiskType, Partition

• Controller.Reserve, MapToGuest, UnmapFromGuest now:

  • add log context fields (host path, partition, reservation)
  • emit debug logs on reserve/map/unmap and slot freed

• disk.Disk.AttachToVM/DetachFromVM/ReservePartition and mount.Mount.MountToGuest/UnmountFromGuest now emit debug/trace logs and enrich context.

Are there functional changes?

1) Yes: panic → error in Mount.MountToGuest when both guests are nil

Old behavior:

• MountToGuest(ctx, nil, nil) panicked.

New behavior:

• it returns an error: both linuxGuest and windowsGuest cannot be nil
• the unit test was changed accordingly (TestMountToGuest_PanicsWhenBothGuestsNil → TestMountToGuest_ErrorWhenBothGuestsNil).

2) Slight semantic tightening in Disk.MountPartitionToGuest when partition wasn’t reserved

Old behavior:

• if partition not in d.mounts, it returned: partition %d not found on disk

New behavior:

• returns a more specific error including controller/lun and “not reserved”.

[jterry75]

Why do you have this and lcow when they have the same build tags?

[helsaawy]

can we follow the same logic in #2633 and have [Linux|Windows]GuestSCSI(Un)Mounter and co be gated by lcow/wcow build tags so the Controller doesn't need fields and logic for both?