We don't go into too much depth about the internals of the workqueue here; in fact, we will merely scratch the surface (as we mentioned previously, our purpose here is to only focus on using the kernel-global workqueue).
It's always recommended that you use the default kernel-global (system) workqueue to consume your asynchronous background work. If this is deemed to be insufficient, don't worry – certain interfaces are exposed that let you create your workqueues. (Keep in mind that doing so will increase stress on the system!) To allocate a new workqueue instance, you can use the alloc_workqueue() API; this is the primary API that's used for creating (allocating) workqueues (via the modern cmwq framework):
include/linux/workqueue.h
struct workqueue_struct *alloc_workqueue(const char *fmt, unsigned int flags, int max_active, ...);
Note that it's exported via EXPORT_SYMBOL_GPL(), which means it's only available...
