linux: full test suite — unit/integration/functional + virtme-ng bench harness
Java simulation tests (unit/):
- Linux0006Test.java: linux-0001 (headerdep 29×) + linux-0006 (btf 500×+) — 4/4 PASS
- LinuxTest.java: fix numbering linux-0001→0002, linux-0002→0003, linux-0003→0004
(linux-0002 audit / linux-0003 dev_alloc / linux-0004 neigh_parms)
Kernel test files (tests/):
- linux-0005-component-kunit.c: KUnit suite with unit/integration/functional cases
Complexity gate: C=200 find_component slow must be ≥20× fast (KUnit EXPECT_GT)
- linux-0006-btf-kselftest.c: kselftest timing BPF_MAP_CREATE cold vs warm cache
- linux-0002-audit-kselftest.sh: auditctl watch + open() timing, F=50 R=20
- linux-0003-0004-net-kselftest.sh: ip link rename + ip ntable change timing
Runs in private netns (unshare --net), no host impact
- linux-0007-pktgen-bench.sh: pktgen proc read timing, 20× gate
- linux-0008-taskstats-kselftest.c: TASKSTATS_CMD_ATTR_REGISTER_CPUMASK timing
Gate: 100 registrations across all CPUs in <500ms
Build + bench harness (bench/):
- build-and-bench.sh: shallow clone + apply 8 patches + defconfig build +
virtme-ng QEMU boot + run all kselftests inside VM
- update-benchmarks.py: parse bench log, write ## Benchmark Results into UNDF posts
Run after bench to update UNDF posts with actual measured ratios
License: all test code GPLv2 (in-kernel), bench scripts public domain
russell@unturf.com
b1e7dd87a1