Re: [PATCH, RFC] reimplement flush_workqueue()

[Andrew Morton]

On Mon, 18 Dec 2006 01:34:16 +0300
Oleg Nesterov <oleg@xxxxxxxxxx> wrote:

> Remove ->remove_sequence, ->insert_sequence, and ->work_done from
> struct cpu_workqueue_struct. To implement flush_workqueue() we can
> queue a barrier work on each CPU and wait for its completition.

Seems sensible.  I seem to recall considering doing it that way when I
initially implemeted flush_workqueue(), but I don't recall why I didn't do
this.  hmm.

> We don't need to worry about CPU going down while we are are sleeping
> on the completition. take_over_work() will move this work on another
> CPU, and the handler will wake up us eventually.
> 
> NOTE: I removed 'int cpu' parameter, flush_workqueue() locks/unlocks
> workqueue_mutex unconditionally. It may be restored, but I think it
> doesn't make much sense, we take the mutex for the very short time,
> and the code becomes simpler.
> 

Taking workqueue_mutex() unconditionally in flush_workqueue() means
that we'll deadlock if a single-threaded workqueue callback handler calls
flush_workqueue().

It's an idiotic thing to do, but I think I spotted a site last week which
does this.  scsi?  Not sure..
-
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/