tree 7d9125d552c84cb612fee03b8bb9e96c5cf56250
parent a4a8d1c3c4ebc52063397d4b7278a569d88f335a
author Surya Garimella <sgarimella@nvidia.com> 1789404219 -0700
committer Surya Garimella <sgarimella@nvidia.com> 1789706253 -0700

meta-hgxr: nsmd/libnsm TAV Mode Index 20 binding

Add libnsm NSM Type 5 Device Mode Index 20 (TAV Mode) codec
support and the nsmd QTM4 handler/D-Bus binding, structurally
mirroring the merged Power Capping Index 27 implementation.

Byte layout (enum8: Default=0/Enabled=1/Disabled=2, 1-byte
payload) mirrors Index 27 structurally and by value assignment.
No written NSM Index 20 contract exists yet -- this is a
best-effort mirror per engineer instruction and must be
verified against a written NSM API contract before this ships.

Changes:
- libnsm/device-configuration.h: DEVICE_MODE_TAV = 20,
  nsm_tav_mode enum, TAV_MODE_DATA_SIZE, with caveat comments.
- libnsm/test/libnsm_device_configuration_test.cpp: tavModeV2
  suite mirroring powerCappingModeV2.
- nsmd/nsmDeviceInventory/nsmSwitch.hpp/.cpp: NsmSwitchTAVMode
  sensor class, toTAVModeFromGet, genRequestMsg,
  handleResponseMsg, setTAVMode, createNsmSwitchTAVMode, gated
  on entity-manager SupportTAVMode in createNsmSwitchDI.
- nsmd/nsmDeviceInventory/test/nsmSwitchFactoryBranch_test.cpp:
  TAVMode test block mirroring PowerCappingMode coverage.
- nsmtool/nsm_config_cmd.cpp: GetTAVMode/SetTAVMode subcommands.

Depends on PDI DGXOPENBMC-28636 (com.nvidia.DeviceMode.TAVMode)
for the generated server.hpp header used by nsmSwitch.hpp.

Review fixups applied:
- Removed internal ticket-ID references and "mirrors power
  capping" style comments from public source/header/test
  comments (kept the technical rationale).
- SupportTAVMode gate now reads via the existing
  dbusPropertyMapAsBool helper instead of a raw
  std::get_if<bool>, so an integer-encoded EM value is
  normalized instead of silently resolving to false.
- createNsmSwitchTAVMode kept inline (not static), matching
  the established file-local factory-helper convention used by
  createNsmSwitchPowerCappingMode and
  createNsmSwitchL1PredictionMode in this file.
- GetTAVMode::parseResponseMsg rejects a missing/short
  CurrentMode instead of silently omitting it, matching the
  GetLTXMode/GetUPhyMode mandatory-CurrentMode convention.
- Do not resolve Default on D-Bus publish: nsmd does not
  resolve a wire Default (0) CurrentMode/PendingMode to a
  guessed Enabled before publishing to D-Bus, on either the
  GET path or after a successful reset-to-default Set. No
  written contract or arch sign-off confirms what TAV's
  default actually is; the design doc only says it comes from
  QM4's own INI configuration, and the identical guess-on-
  mirror pattern was already wrong once for UPhy mode (commit
  c7c3b2b4). nsmd now publishes the wire value as-is; bmcweb's
  TAVMode handlers already treat a still-Default
  CurrentMode/PendingMode as not-yet-resolved by firmware,
  matching the existing LTXMode/UPhyRecoveryMode contract.

Note: the nsmtool GetTAVMode CLI display intentionally still
prints raw "Default" for wire value 0 (not "Enabled"),
matching the established GetPowerCappingMode/GetLTXMode/
GetUPhyMode CLI precedent in this file. A prior review
comment suggested mapping Default to Enabled in the CLI to
match D-Bus read semantics, but that no longer applies now
that nsmd does not resolve Default to Enabled on the D-Bus
side either -- see the note above. Flagging for tech-lead
call rather than reintroducing the guess.

Test status: local build blocked in this environment by a
missing dbus-1 pkg-config dependency (nsmtool/meson.build);
clang-format-20 clean on all changed files. No lab/unit run
performed -- pending CI and PDI MR merge.

Fixes jira https://jirasw.nvidia.com/browse/DGXOPENBMC-28638

Signed-off-by: Surya Garimella <sgarimella@nvidia.com>