On Tue, 2018-01-23 at 09:42 -0800, Linus Torvalds wrote:
> On Tue, Jan 23, 2018 at 8:57 AM, Paolo Abeni <pab...@redhat.com> wrote:
> >
> >> Or is it that the workqueue execution is simply not yielding for some
> >> reason?
> >
> > It's like that.
> >
> > I spent little time on it, so I haven't many data point. I'll try to
> > investigate the scenario later this week.
>
> Hmm. workqueues seem to use cond_resched_rcu_qs(), which does a
> cond_resched() (and a RCU quiescent note).
>
> But I wonder if the test triggers the "lets run lots of workqueue
> threads", and then the single-threaded user space just gets blown out
> of the water by many kernel threads. Each thread gets its own "fair"
> amount of CPU, but..
If folks aren't careful with workqueues, they can be a generic
starvation problem. Like the below in the here and now.
fs/nfs: Add a resched point to nfs_commit_release_pages()
During heavy NFS write, kworkers can do very large amounts of work
without scheduling (82ms traced). Add a resched point.
Signed-off-by: Mike Galbraith <efa...@gmx.de>
Suggested-by: Trond Myklebust <tron...@primarydata.com>
---
fs/nfs/write.c | 1 +
1 file changed, 1 insertion(+)
--- a/fs/nfs/write.c
+++ b/fs/nfs/write.c
@@ -1837,6 +1837,7 @@ static void nfs_commit_release_pages(str
set_bit(NFS_CONTEXT_RESEND_WRITES, &req->wb_context->flags);
next:
nfs_unlock_and_release_request(req);
+ cond_resched();
}
nfss = NFS_SERVER(data->inode);
if (atomic_long_read(&nfss->writeback) < NFS_CONGESTION_OFF_THRESH)
