Date: Sat, 28 Oct 2006 21:57:10 -0700 From: Paul Saab <ps@freebsd.org> To: Mike Jakubik <mikej@rogers.com> Cc: hardware@freebsd.org Subject: Re: GIANT-FREE ciss Message-ID: <454434A6.8030207@freebsd.org> In-Reply-To: <4542AFFD.7070100@rogers.com> References: <45429B67.4020500@rogers.com> <4542A948.5080809@freebsd.org> <4542AFFD.7070100@rogers.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Mike Jakubik wrote: > Paul Saab wrote: >> I did experiments and found that ciss did not gain any thing by >> making it MPSAFE. This may have changed with newer cards, but with >> all the older cards, there was no gain. > > Thanks for the information. I wasn't aware that this was a possible > case. Wouldn't the GIANT lock have a negative effect on other > applications/system? Or is my understanding of the lock flawed? In all the tests I did, there was no improvement with a non-GIANT ciss. There are a few problems with getting ciss out from under giant. 1. CAM needs to be out from under giant. Until this is done, you can't make the current incarnation of ciss GIANT free 2. ciss could be converted to a block driver but then this would break people since it would introduce /dev/cissd, but when I did the experiments to convert the driver to this method, I again did not see any gains vs the CAM driver, so I tabled the work.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?454434A6.8030207>