netdev - Re: [PATCH net-next 2/2] sunvnet: Re-check for a VIO_DESC_READY data descriptor after short udelay()

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: Windows password security audit tool. GUI, reports in PDF.

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

Date:	Tue, 2 Sep 2014 10:33:07 -0700
From:	Raghuram Kothakota <Raghuram.Kothakota@...cle.com>
To:	Sowmini Varadhan <sowmini.varadhan@...cle.com>
Cc:	David Miller <davem@...emloft.net>, netdev@...r.kernel.org
Subject: Re: [PATCH net-next 2/2] sunvnet: Re-check for a VIO_DESC_READY data descriptor after short udelay()


On Sep 2, 2014, at 9:56 AM, Sowmini Varadhan <sowmini.varadhan@...cle.com> wrote:

> On (09/02/14 09:43), Raghuram Kothakota wrote:
>> We could optimize this a bit by not wait for normal traffic, I mean,
>> non-burst traffic
>> and apply the retry  only  when we detect a stream of packets?
> 
> How could you tell the difference efficiently? You'd need to
> track some kind of history/state for inter-packet arrival time.
> All seems like over-kill (more useful to go and optimize other
> parts of the system, such as do less work in interrupt context).

>From what I see, the vnet_walk_rx() picks up packets in a while loop,
we could count the number of packets picked up in that loop and use
that count as a method to determine if we need to apply this retry or not.
That is, retry only if that counter is > x,  that may avoid waiting for cases
where peer sent one packet only?   It may be worth trying it
and see if it still keeps up the improvement that you saw.

-Raghuram
> 
>> We could detect a stream based on how many packets are picked
>> up in this function, picking up 3 or more could be considered as a stream,
>> of course tune based on testing.
>> 
>> You probably tried it already, but checking to see if you tried with
>> less number
>> of iterations, we could reduce the iterations if the numbers are equally good
>> with less iterations.
> 
> yes, the 3 * 4 micro-seconds was arrived at heuristically.
> 
> --Sowmini
> 

--
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