Re: [PATCH 0/3] workqueue: Add telemetry tracepoints for CPU hogs, distress, and BH budget yields

[Tejun Heo]

Hello, Aaron.

On Sat, Aug 29, 2026 at 07:05:14PM -0400, Aaron Tomlin wrote:
> This patch series introduces lightweight tracepoints for these key
> operational boundaries:

Lai's address was mangled in the cc list. Corrected to
jiangshanlai@xxxxxxxxx.

Generally looks fine to me. Some comments:

- As the test robot reported, the open-coded u64 division in the first
  patch breaks 32bit builds. Rather than restructuring the comparison,
  it'd be better to keep it as-is and calculate the duration only after
  the worker is marked CPU_INTENSIVE. That also keeps the division out
  of the every-tick path.

- In the third patch, the timeout flag is determined by re-reading
  jiffies after the loop. If the loop exited because nr_restarts ran
  out, time_before() was never tested and a tick in that window would
  misattribute the yield to timeout. Please derive the reason from the
  condition that actually terminated the loop. Also, BH_WORKER_RESTARTS
  - nr_restarts counts loop iterations, not restarts.

Thanks.

-- 
tejun