[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <4714EC54.7080808@garzik.org>
Date: Tue, 16 Oct 2007 12:52:36 -0400
From: Jeff Garzik <jeff@...zik.org>
To: Ingo Molnar <mingo@...e.hu>
CC: David Miller <davem@...emloft.net>, linux-kernel@...r.kernel.org,
torvalds@...ux-foundation.org, shemminger@...ux-foundation.org
Subject: Re: [patch] forcedeth: fix the NAPI poll function
Ingo Molnar wrote:
> * Jeff Garzik <jeff@...zik.org> wrote:
>
>>> FYI, looks good so far, ontop of Linus' latest -git tree.
>>>
>>> btw., one thing i always found weird about forcedeth is that it
>>> generates an extra ~100 irqs per second even when there's no network
>>> traffic:
>>>
>>> 11: 89752 XT-PIC-XT eth0
>>> 11: 89854 XT-PIC-XT eth0
>>> 11: 89955 XT-PIC-XT eth0
>>> 11: 90056 XT-PIC-XT eth0
>>> 11: 90157 XT-PIC-XT eth0
>>>
>>> (irq count snapshot every second). Just in case it's easy to fix ...
>>
>> May I presume that behavior occurs in both mainline and my fe-lock
>> branch?
>
> correct.
Any chance of getting a quick comparison of forcedeth CPU usage in your
setup, between old driver and fe-lock driver?
I'm curious about the effects of fe-lock on CPU usage, which NAPI is
known to perturb (in positive or negative directions, depending on
conditions).
I presume both drivers can work at wire speed, sans any "too many
iterations" problems...
>> That is likely the NIC's timer irq, which is programmed to kick the
>> irq handler every so often. DEV_NEED_TIMERIRQ is a starting point, as
>> well as "Known bugs" at the top of forcedeth.c, if you're curious.
>>
>> I'm _betting_ that some of that can be eliminated, but we'll see...
>
> ok. Was just a FYI - you were looking for feedback :-)
Thanks! :)
Jeff
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists