fastrpc-test: remove SoC blacklists and SM8850 exception

[anankulk]

Summary

Remove hardcoded SoC-based domain filtering and per-SoC PD support maps.
Previously blacklisted platforms (e.g. Glymur CRD) were fully skipped
regardless of actual hardware capability. This PR replaces that approach
with runtime discovery driven by fastrpc-healthcheck.

Changes

• Use fastrpc-healthcheck as the primary capability source for domain
  online state and signed/unsigned PD support per domain
• An installed healthcheck that fails, times out, or produces malformed
  output fails the suite; runtime fallback applies only when tool is absent
• Fall back to remoteproc running state and /dev/fastrpc-<domain>[-secure]
  endpoint presence for domain discovery when healthcheck is unavailable
• PD support in fallback mode uses a conservative static protocol map
  (ADSP/MDSP/SDSP: signed only; CDSP/CDSP1/GDSP0/GDSP1: signed + unsigned)
• Add domain_source and pd_source fields to [FASTRPC-POLICY] log line
• Automatic mode skips domains missing runtime readiness, endpoint, or
  library; explicit selection fails with a diagnostic message
• Strict healthcheck parser with field validation, duplicate detection, and
  GPDSP0/GPDSP1 canonicalization
• Identity-based remoteproc-to-domain mapping replacing fragile basename
  guessing; FASTRPC_DOMAIN/FASTRPC_DOMAIN_NAME force single-domain
  independently of --domain-mode
• Add --healthcheck-timeout parameter (default 15s)
• Enforce finite --timeout with explicit default of 120s
• Validate option arguments and FASTRPC_UNSIGNED_PD environment variable
• Discover DSP skeleton directories from actual .so presence, not hardcoded ABI names
• Capture kernel evidence once after the run matrix on any failure

Validation

Platform	Capability Source	Domains	Invocations	Result
QCS9075	runtime-fallback	ADSP/CDSP/GDSP0/GDSP1	9/9	PASS
QCS9075	fastrpc-healthcheck	ADSP/CDSP/GDSP0/GDSP1	9/9	PASS
Glymur CRD	fastrpc-healthcheck	ADSP/CDSP	3/3	PASS

[Srikanth Muppandam (smuppand)]

only_hap_example_failed() now has no callers and still documents the SM8850 exception being removed. Remove the function and obsolete comment in this PR.

[anankulk]

Hi Srikanth Muppandam (@smuppand),

With the upcoming tag to be released for fastrpc recipe : we are enabling gpdsp test support as-well.
Once the tag is released, will address the comments and enable gdsp as-well along with the on-going changes.

[Srikanth Muppandam (smuppand)]

With these new changes, you have to fix the following code as well.

Runner/utils/lib_fastrpc.sh:151 — artifact discovery validates generic directories, not FastRPC artifacts

• Issue: FASTRPC_RESOLVED_LIB_SYS_DIR becomes /usr/lib whenever that directory exists. Likewise, the runner checks FASTRPC_RESOLVED_SKEL_BASE,
  although the base can exist without either v75 or v68. FASTRPC_RESOLVED_LIB_TEST_DIR is discovered but never required.

• Recommended fix: Resolve and validate the actual required library files, require a non-empty FASTRPC_RESOLVED_SKEL_PATH, and validate the required
  test-library artifacts rather than generic directories.

Runner/suites/Multimedia/CDSP/fastrpc_test/fastrpc_test.yaml:11 — YAML still documents removed GPDSP filtering

• Issue: The description says GPDSP domains are skipped on QCS9075/QCS8275/QCS8300/QCS9100, while the PR removes that blacklist.
• Recommended fix: Describe runtime domain and endpoint discovery instead.

Runner/suites/Multimedia/CDSP/fastrpc_test/run.sh:188 — remove unused SOC_MACHINE

• Recommended fix: Remove the assignment.

[Srikanth Muppandam (smuppand)]

anankulk Please address the pending comments too

[Srikanth Muppandam (smuppand)]

anankulk sign-off is missing from the recent patches.

[Srikanth Muppandam (smuppand)]

anankulk Restore executable mode on run.sh.

[Srikanth Muppandam (smuppand)]

Fixes #586

[Srikanth Muppandam (smuppand)]

This PR looks ready to merge. Once the latest comment is addressed, we should be good to go.

[Srikanth Muppandam (smuppand)]

The updated suite still writes results directly through repeated echo "$TESTNAME : ..." paths and maintains separate PASS/FAIL counters instead of using test_result_init, test_result_record, and test_result_finish.

Because the result file is not cleared at startup, an interrupted or unexpectedly terminated run can leave a previous PASS result for send-to-lava.sh to consume. The many newly added early exits also duplicate finalization behavior.

Initialize the result state immediately after defining TESTNAME and RES_FILE, record invocation outcomes through test_result_record, and route every terminal PASS/FAIL/SKIP path through test_result_finish.

[anankulk]

Acknowledged.

• test_result_init "$TESTNAME" "$RESULT_FILE" called immediately after RESULT_FILE is defined — clears any stale result file at startup
• All per-invocation outcomes routed through test_result_record "PASS"/"FAIL" — separate PASS_COUNT/TOTAL_COUNT variables removed
• Every terminal path (early validation, healthcheck failures, domain selection, final result) routed through test_result_finish
• Validated on QCS9075: 9 invocations, 9 passed

[Srikanth Muppandam (smuppand)]

LGTM