Date: Thu, 7 Dec 2006 15:22:54 +0100 From: Paolo Pisati <piso@freebsd.org> To: Andre Oppermann <andre@freebsd.org> Cc: Gleb Smirnoff <glebius@freebsd.org>, Perforce Change Reviews <perforce@freebsd.org>, Paolo Pisati <piso@freebsd.org> Subject: Re: PERFORCE change 111230 for review Message-ID: <20061207142254.GA1195@tin.it> In-Reply-To: <4578070A.2030609@freebsd.org> References: <200612062319.kB6NJgsq031755@repoman.freebsd.org> <20061207110225.GU32700@FreeBSD.org> <4578070A.2030609@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Dec 07, 2006 at 01:20:26PM +0100, Andre Oppermann wrote: > Gleb Smirnoff wrote: > >On Wed, Dec 06, 2006 at 11:19:42PM +0000, Paolo Pisati wrote: > >P> http://perforce.freebsd.org/chv.cgi?CH=111230 > >P> > >P> Change 111230 by piso@piso_newluxor on 2006/12/06 23:19:06 > >P> > >P> To workaround a problem when an mbuf is bigger than MCLBYTES > >P> (and thus not handled by m_megapullup()), introduce & use > >P> m_jumbo16pullup() that uses 16k jumbo size cluster. > >P> > >P> This problem only showed up on traffic generated on a box with > >P> a nic tso enabled. > > > >Paolo, > > > >this isn't a fix. Another application will do write(,, 16k + 1) and > >m_jumbo16pullup() will fail again. Please backout it, it is a hack. > > > >We need to fix TSO in such way that real packets, that will be > >transmitted to wire, will be passed to pfil handlers. > > That is not possible. Then, after a discussion on irc, pull the plug on any present (and future) hackery & half-baked solution, and declare in kernel libalias incompatible with tso. bye -- Paolo Piso's first law: nothing works as expected!
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20061207142254.GA1195>