Re: if_wm.c workqueue

To: Kengo NAKAHARA <k-nakahara%iij.ad.jp@localhost>
Subject: Re: if_wm.c workqueue
From: Emmanuel Dreyfus <manu%netbsd.org@localhost>
Date: Thu, 26 May 2022 02:20:46 +0000

On Thu, May 26, 2022 at 10:28:03AM +0900, Kengo NAKAHARA wrote:
> When hw.wm0.txrx_workqueue is set to 1, receive and transmit processing
> of wm0 uses workqueue(9) instead of softint(9), that is, the processing
> can be preempted.  As a result, userland process can run even if wm0
> receives packets at very high rate.  However, that may causes degradation
> of throughput and latency.

Right, the behavior seems desirable, however the thing seems to have
rotten. The interface will not work once enabled.

-- 
Emmanuel Dreyfus
manu%netbsd.org@localhost

Follow-Ups:
- Re: if_wm.c workqueue
  - From: Kengo NAKAHARA

References:
- if_wm.c workqueue
  - From: Emmanuel Dreyfus
- Re: if_wm.c workqueue
  - From: Kengo NAKAHARA

Prev by Date: Re: if_wm.c workqueue
Next by Date: Re: if_wm.c workqueue
Previous by Thread: Re: if_wm.c workqueue
Next by Thread: Re: if_wm.c workqueue
Indexes:

Home | Main Index | Thread Index | Old Index