nx_ctxtop.nx

nx_ctxtop.nx -- WHO is producing the context switches. The producer-finder for seq1340. nx_procchurn measured the box at 81k-93k ctxsw/sec in every sample (the one CONSTANT in the perf lane) and DISPROVED forks as the driver -- forks swing 15->179->10 while ctxsw stays flat. A box-level rate names no culprit, so this walks /proc/<pid>/status twice and attributes the switches per process. LAW (rising-debt-means-find-the-producer): a rate with no owner is a symptom, not a diagnosis. ★Reports voluntary and nonvoluntary SEPARATELY because they demand OPPOSITE fixes: voluntary = the process blocked itself (poll/read/sleep) -> batch, lengthen polls, go event-driven; nonvoluntary = the scheduler preempted it -> reduce concurrency. A tool printing only the sum sends the fix the wrong way. Exit code = 0 when the sample completed, 3 UNMEASURED (fail-closed, never a confident empty answer). ★2026-08-12 -- PRODUCER-CLASS ATTRIBUTION: the ranked rows name WHO; a final NX-CTXTOP-CLASS line names WHAT KIND (SYSTEM=DSM/Synology, ESTATE=nx_* organs, KERNEL=kthreads, OTHER) so a RED churn axis carries its own verdict instead of a human minting a cap and eyeballing comm strings. Field-measured 2026-08-12: every top-8 producer was DSM (ssrtmpserverd/synorelayd/syno-cloud-clie/redis/kworkers), zero estate. license_tier: ORIGINAL Read-only. No hw writes (Rule 26). expect_exit: 0