On Wed, 24 Sept 2025 at 15:06, Michal Pecio [off-list ref] wrote:

I think yes, usually in USB-speak "completion" is when the URB is
finished for any reason, including error or unlink/cancellation.
"Free" could suggest usb_free_urb().

But I see your point. Maybe "finish execution" is less ambiguous?

I will use completion if it's the standard terminology

I think it's an irrelevant detail which CPU executed which function.
It could all happen sequentially on a single core and it's still the
same bug.

In fact, I just reproduced it with all CPUs offlined except one.

My bad, I see it now. I keep forgetting the actual urb execution
is asynchronous

Thanks
Viswanath