Date: Mon, 10 May 2010 15:07:46 -0400 From: Sam Leffler <sam@errno.com> To: Luigi Rizzo <rizzo@iet.unipi.it> Cc: net@freebsd.org Subject: Re: max_linkhdr defaults to 16, too short ? Message-ID: <DE43B7BC-2208-4DC5-B7FD-C6A246E1EA8C@errno.com> In-Reply-To: <20100420215527.GA75324@onelab2.iet.unipi.it> References: <20100420172845.GA71187@onelab2.iet.unipi.it> <20100420215527.GA75324@onelab2.iet.unipi.it>
next in thread | previous in thread | raw e-mail | index | archive | help
On Apr 20, 2010, at 5:55 PM, Luigi Rizzo wrote: > On Tue, Apr 20, 2010 at 07:28:45PM +0200, Luigi Rizzo wrote: >> just noticed that sys/kern/uipc_domain.c still sets max_linkhdr=16 >> as a default. >> The value is often used to reserve head space in mbufs for >> the MAC header. As such, 16 is too short for systems that make >> use of vlans, and the effect might be that we would need >> additional mbuf entries or at least move stuff down >> as the vlan tag is added. >> >> Any objection to bumping the default to 20 ? > > forgot to mention: > > max_linkhdr is available as a sysctl, kern.max_linkhdr , but other > than that, there is no code in sys/ that sets the value, so systems > are stuck at 16 unless users override the default. We need a coherent way to handle max_linkhdr. I hacked it in net80211 to insure bridged 802.11 frames have sufficient contiguous space in the mbufs but never did something like define an api and generate events/callbacks on changes. Last I looked doing this right was non- trivial. Sam
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?DE43B7BC-2208-4DC5-B7FD-C6A246E1EA8C>