Re: [patch] generic work queue handling, workqueue-2.5.39-D6

[Ingo Molnar]

On Mon, 30 Sep 2002, Dipankar Sarma wrote:

> Ingo,
> 
> Is it possible that queue_task() handlers in earlier driver code may
> have depended on implicit serialization against corresponding timer
> handlers since each of those is run from BHs ? If so, isn't that an
> issue now with no BHs ? Or, is it safe to assume that general smp-safety
> code in the drivers will take care of serialization between timers and
> work-queues ?

yes, this is true - such drivers need to use spinlocks. But since 
basically every driver abstraction within the kernel already necessiates 
per-driver spinlocks, this should be straightforward in most cases.

	Ingo