Re: Workqueue behaviour - Synchronization within submitted work.

[Stefan Richter]

On Mar 20 Vijay wrote:
> In the new workqueue architecture, I have a question regarding
> synchronization between different "work" submitted to the same
> workqueue. For example:
> 
> * I submit two "works" A and B to a common driver specific workqueue (W).
> * Each A and B, meddle with a certain shared data SD.
> * Am I guaranteed serialized execution of A and B, or should I play
> safe and use a semaphore (acquire/release) with A and B ?

The works A and B may be (and more or less likely will be) executed in
parallel --- except if you create the workqueue with parameters @flags |=
WQ_UNBOUND and @max_active = 1.  alloc_ordered_workqueue() is a shorthand
for creation of such a workqueue.

Some more thoughts:
  - If you go for an ordered workqueue and no other means of serialization
    of accesses to your data, check carefully that these data are indeed
    never accessed in other contexts besides the workers.
  - For serialization in non-atomic context, a struct mutex is usually
    preferred over a struct semaphore.  The foremost advantage of mutexes
    over counting semaphores is that code using the former can be debugged
    much easier.
-- 
Stefan Richter
-=====-===-= --== ==---
http://arcgraph.de/sr/
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/