Date: Fri, 1 Nov 1996 14:35:35 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: matt@lkg.dec.com (Matt Thomas) Cc: terry@lambert.org, msmith@atrad.adelaide.edu.au, freebsd-hackers@freefall.freebsd.org Subject: Re: VM answer requested Message-ID: <199611012135.OAA28514@phaeton.artisoft.com> In-Reply-To: <199611011656.QAA14425@whydos.lkg.dec.com> from "Matt Thomas" at Nov 1, 96 04:56:00 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > What about the aligned access bit in recent Intel processors? > > > > I'd like to be able to turn on the bit so that I get a fault (and it > > kills the offending process) when an unaligned access occurs. > > > > I'd also like to get kernel faults if this happens in the kernel > > (it is my opinion that it should never be allowed to happen in the > > kernel, and there should be a sanitization pass to insure it). > > Don't use the de driver then. Because of a misfeature in the 21x4x > chips, the Ethernet payload (the stuff after the header) is not longword > aligned). Having the processor deal with unaligned data is much faster > than copying the data so that it is aligned. A "relaxed but observant" mode would handle this in the fault handler by falling back to unaligned access -- after the fault. I expect the kernel would run "relaxed but observant" and the user mode code would be "anal for debugging, relaxed-but-observant for normal usage". How will this driver work with an Alpha or MIPs processor, BTW? It's a DEC card, you'd expect it to work with both... Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199611012135.OAA28514>