Date: Fri, 24 Jan 2014 15:05:57 -0800 From: John-Mark Gurney <jmg@funkthat.com> To: Rick Macklem <rmacklem@uoguelph.ca> Cc: freebsd-net@freebsd.org, J David <j.david.lists@gmail.com> Subject: Re: Terrible NFS performance under 9.2-RELEASE? Message-ID: <20140124230557.GF75135@funkthat.com> In-Reply-To: <659117348.16015750.1390604069888.JavaMail.root@uoguelph.ca> References: <CABXB=RSebaWTD1LjQz__ZZ3EJwTpOMpxq0Q=bt4280dx%2B0auCw@mail.gmail.com> <659117348.16015750.1390604069888.JavaMail.root@uoguelph.ca>
next in thread | previous in thread | raw e-mail | index | archive | help
Rick Macklem wrote this message on Fri, Jan 24, 2014 at 17:54 -0500:
> The largest jumbo packet supported by the generic mbuf code is 16K
> (or maybe 9K for 9.2). I have no idea if this matters or not.
This is only partly true. Our allocators only supports mbufs of
2k (standard size), 4k (page size), 9216 and 16184... If you allocate
a 9k or 16k mbuf, it is guaranteed that the data will be physically
contiguous so that cards that can't do scatter/gather DMA can handle
larger frames... But if the card can handle S/G DMA, they can send a
32KB packet made up of normal 2k clusters, or any other odd sized
mbufs...
There are only a couple drivers (and I plan on working to remove the
limit) that limit the size of MTU, but if the driver hardware supports
it, there is nothing in our stack preventing large, as in 64KB, MTU
use..
--
John-Mark Gurney Voice: +1 415 225 5579
"All that I will do, has been done, All that I have, has not."
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20140124230557.GF75135>