Re: [PATCH rfc] workqueue: honour cond_resched() more effectively.

[Trond Myklebust <trondmy@hammerspace.com>]

On Mon, 2020-11-09 at 09:01 -0500, tj@kernel.org wrote:
> Hello,
> 
> On Mon, Nov 09, 2020 at 01:50:40PM +0000, Trond Myklebust wrote:
> > > I'm thinking the real problem is that you're abusing workqueues.
> > > Just
> > > don't stuff so much work into it that this becomes a problem. Or
> > > rather,
> > > if you do, don't lie to it about it.
> > 
> > If we can't use workqueues to call iput_final() on an inode, then
> > what
> > is the point of having them at all?
> > 
> > Neil's use case is simply a file that has managed to accumulate a
> > seriously large page cache, and is therefore taking a long time to
> > complete the call to truncate_inode_pages_final(). Are you saying
> > we
> > have to allocate a dedicated thread for every case where this
> > happens?
> 
> I think the right thing to do here is setting CPU_INTENSIVE or using
> an
> unbound workqueue. Concurrency controlled per-cpu workqueue is
> unlikely to
> be a good fit if the work can run long enough to need cond_resched().
> Better
> to let the scheduler handle it. Making workqueue warn against long-
> running
> concurrency managed per-cpu work items would be great. I'll put that
> on my
> todo list but if anyone is interested please be my guest.
> 

That means changing all filesystem code to use cpu-intensive queues. As
far as I can tell, they all use workqueues (most of them using the
standard system queue) for fput(), dput() and/or iput() calls.

-- 
Trond Myklebust
Linux NFS client maintainer, Hammerspace
trond.myklebust@hammerspace.com