Re: [PATCH 2/3] workqueue: expose per-workqueue stats in debugfs

[Wilson Felipe Pereira]

Tejun Heo <tj@xxxxxxxxxx> writes:

> On Thu, Oct 22, 2026 at 03:53:35PM +0000, Wilson Felipe Pereira wrote:
>> struct pool_workqueue tracks runtime counters in pwq->stats[],
>> pwq->total_in_flight, and pwq->nr_active, but reading them currently
>> requires tools/workqueue/wq_monitor.py with drgn and kernel DWARF
>> debuginfo. This makes it impractical to inspect workqueue stats on
>> systems without drgn or debug symbols.
>>
>> Add /sys/kernel/debug/workqueue/stats to expose per-workqueue stats
>> directly from the kernel. Iterate over the workqueues list under RCU and
>> sum wq->retired_stats[] with each pwq's stats[], total_in_flight, and
>> nr_active to report queued, active, and running work items alongside the
>> cumulative pwq->stats[] counters.
>
> This isn't really a justification. What do you use these numbers for? If
> this is for live monitoring, why debugfs? If for debugging, what's wrong
> with drgn?

This will mainly be used for continuous monitoring and exporting the
stats from the machine (CPU time, queued backlog, etc.). Also to quickly
check if any item is queued or stuck.

Why not drgn? For continuous monitoring, running drgn on every machine,
with vmlinux debuginfo, and memory footprint from Python/drgn, while
having CONFIG_PROC_KCORE available, it adds up. Having a simple file
interface, it can be read from an existing monitoring collector.

Why debugfs? If sysfs is picked, not all queues are in
/sys/bus/workqueue/devices/. Using debugfs, a single file can include
all queues. If you have a suggestion, I'm open to change this series.

Thanks!