DPCs

DPCs, or Deferred Procedure Calls, are a kernel mechanism used by the Windows operating system to defer work that cannot be safely performed at an interrupt level. When a device interrupt occurs, the corresponding interrupt service routine (ISR) runs at a high interrupt request level (IRQL) and handles only time-critical tasks. The ISR often queues a DPC for later processing, allowing more extensive work to be completed outside the immediate interrupt context.

A DPC is represented by a small object containing a pointer to a routine (the DeferredRoutine) and several context fields (DeferredContext, SystemArg1, SystemArg2). The DPC dispatcher processes these queued entries on the processor at the DISPATCH_LEVEL IRQL. In practice, DPCs are used to complete I/O processing, update data structures, or schedule work that would block or take too long at a higher IRQL.

DPCs are typically per-processor to reduce cross-CPU locking and to improve latency. They are inserted with the kernel function KeInsertQueueDpc and can be removed with KeRemoveQueueDpc. DPC routines must be non-blocking and should avoid operations that may block or wait for resources; they should be short and finish quickly to prevent excessive DPC latency, which can degrade system responsiveness.

In driver development, DPCs are a common mechanism to separate fast interrupt handling from slower processing, enabling efficient I/O completion paths. Developers monitor DPC latency and optimize DPC routines to minimize delays, as high latency can impact real-time performance in systems with stringent timing requirements.

Openly, DPC can refer to other concepts in different domains, but this article describes Deferred Procedure Calls in the Windows kernel.