netdev - Re: [PATCH net-next] net: allow large number of tx queues

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]

Message-ID: <20130623102951.GC17316@redhat.com>
Date:	Sun, 23 Jun 2013 13:29:51 +0300
From:	"Michael S. Tsirkin" <mst@...hat.com>
To:	Eric Dumazet <eric.dumazet@...il.com>
Cc:	Jason Wang <jasowang@...hat.com>, davem@...emloft.net,
	hkchu@...gle.com, netdev <netdev@...r.kernel.org>,
	dwmw2@...radead.org
Subject: Re: [PATCH net-next] net: allow large number of tx queues

On Fri, Jun 21, 2013 at 12:12:30AM -0700, Eric Dumazet wrote:
> On Fri, 2013-06-21 at 14:41 +0800, Jason Wang wrote:
> 
> > 1K queues (about 80 pages) looks a little bit aggressive which means we
> > may always fall back to vmalloc()?
> 
> 1K queues looks like you should use 0 queue (LLTX) in fact, so that you
> design the minimal percpu structure in the driver to fit the needs.
> 
> Using Qdisc, qith MQ, pfifo_fast, on 1000 'queues' is adding a fair
> amount of overhead and memory usage.
> 
> For example, loopback driver only needs small percpu structure (16
> bytes), as it immediately forward the packet into netif_rx() [ Which
> also use a percpu area called softnet_data ]
> 
> 

I agree it's worth considering for tun.

But the same number of queues will be there for virtio net,
and that's much more like a regular network device.

-- 
MST
--
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