[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1487172397.1311.23.camel@edumazet-glaptop3.roam.corp.google.com>
Date: Wed, 15 Feb 2017 07:26:37 -0800
From: Eric Dumazet <eric.dumazet@...il.com>
To: "Matan Barak (External)" <matanb@...lanox.com>
Cc: Saeed Mahameed <saeedm@....mellanox.co.il>,
David Miller <davem@...emloft.net>,
netdev <netdev@...r.kernel.org>,
Tariq Toukan <tariqt@...lanox.com>,
Saeed Mahameed <saeedm@...lanox.com>, jackm@...lanox.com
Subject: Re: [PATCH net-next] mlx4: do not fire tasklet unless necessary
On Wed, 2017-02-15 at 16:52 +0200, Matan Barak (External) wrote:
> So, in case of RDMA CQs, we add some per-CQE overhead of comparing the
> list pointers and condition upon that. Maybe we could add an
> invoke_tasklet boolean field on mlx4_cq and return its value from
> mlx4_cq_completion.
> That's way we could do invoke_tasklet |= mlx4_cq_completion(....);
>
> Outside the while loop we could just
> if (invoke_tasklet)
> tasklet_schedule
>
> Anyway, I guess that even with per-CQE overhead, the performance impact
> here is pretty negligible - so I guess that's fine too :)
Real question or suggestion would be to use/fire a tasklet only under
stress.
Firing a tasklet adds a lot of latencies for user-space CQ completion,
since softirqs might have to be handled by a kernel thread (ksoftirqd)
I would be surprised if no customer was hit by your commit,
( net/mlx4_core: Use tasklet for user-space CQ completion events )
especially when using specific (RT) scheduler classes.
Powered by blists - more mailing lists