Re: Races in power IRP handling path.

[Paul Durrant <xadimgnik@xxxxxxxxx>]

On 11/10/2022 15:33, Martin Harvey wrote:
> Hi,
>
> Been debugging power IRP handling.
Which driver?

> I notice that we not only pend power 
> IRPs on the dispatch path, but also use completion routines to try to 
> force a synchronous completion (Forward irp synchronously). You might 
> want to consider that setting a completion routine **outside** the 
> dispatch handler on a pended IRP is likely to be a racey operation in 
> and of itself.
Racey in what way? Power IRPs are special... there can only be one 
outstanding.
  Paul