On Wed, 2 Sep 2026 06:54:32 +0000 [email protected] wrote: > Could the commit message be updated to reflect that the buffers are not > cleared?
> Does this code unnecessarily export tracing_reset_all_cpus() outside of > trace.c? Both are real. Fixed in v6. The commit message was left over from the stricter reset design that was dropped in favour of map-only reset. It no longer claims the trace_array buffers are cleared. ftrace_stackmap_create() still takes the owning trace_array, but for a different reason: the text and binary exports run stack addresses through trace_adjust_address(). The export was indeed dead. With the ring-buffer clear gone, tracing_reset_all_cpus() has no caller outside trace.c, so v6 restores it to a private static helper and removes the declaration from trace.h. Pengfei
