netdev - Re: [PATCH v2.6.29 v3 5/5] gianfar: Continue polling until both tx and rx are empty

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Wed, 17 Dec 2008 20:40:20 -0600
From:	"Fleming Andy-AFLEMING" <afleming@...escale.com>
To:	"David Miller" <davem@...emloft.net>
Cc:	"Jeff Garzik" <jeff@...zik.org>, <netdev@...r.kernel.org>
Subject: Re: [PATCH v2.6.29 v3 5/5] gianfar: Continue polling until both tx and rx are empty



On Dec 17, 2008, at 18:55, "David Miller" <davem@...emloft.net> wrote:

> From: Andy Fleming <afleming@...escale.com>
> Date: Wed, 17 Dec 2008 15:42:55 -0600
>
>> gfar_poll would declare polling done once the rx queue was empty,
>> but the tx queue could still have packets left.
>>
>> Stolen mostly from the e1000 driver.
>>
>> Signed-off-by: Andy Fleming <afleming@...escale.com>
>
> Applied.
>
> But I absolutely do not recommend this polling technique at all.
>
> The best scheme is to first purge your TX ring completely, and do not
> apply any quota to this work.  It's relatively cheap and batches well.
>
> Then, you process the RX ring and apply the quota only to the RX ring
> work.
>
> This is what every NAPI driver I've written does and it gives the best
> results.

I agree it seems a bit screwy, but I got a 4-5% performance increase.   
It may just require better tuning of my tx coalescing parameters-I was  
getting 7x the number of tx irqs as rx.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html