Re: [PATCH 0/4] Change return values from queue_work et al.

[Stefan Richter <stefanr@s5r6.in-berlin.de>]

Alan Stern wrote:
...
> Note that the change falls within the bounds of the documented
> behavior, in the sense that any code which was originally written
> correctly (i.e., in accordance with the documentation) will continue to
> work correctly without generating any warnings.
...

You are right that there is no comment (or better yet, kerneldoc
comment) about what happens if an instance of work_struct is enqueued
twice. However, /a/ there is the source and /b/ Corbet, Rubini,
Kroah-Hartman: LDD3 describes in detail in an easily understood section
how workqueues are to be used. (Workqueues in Linux 2.6.10, that is.)

...
> If the
> usage is correct then there is no harm in leaving the WARN_ON call where
> it is.  If the usage is wrong then the call needs to be fixed, and the
> maintainer for the subsystem containing the call will soon find out about
> it, thanks to the WARN_ON.
...

Acceptable on second thought, particularly in light of your new
replacement functions with improved semantics of their return value.
Although there are cases where the WARN_ON might not go off during a
long time, or where an update won't happen in many months despite
hundreds of reports at dozens of mailing lists and bugzillas.
-- 
Stefan Richter
-=====-=-==- =--- ===-=
http://arcgraph.de/sr/