Date: Fri, 20 Jul 2012 00:27:52 -0700 From: Adrian Chadd <adrian@freebsd.org> To: mj@feral.com Cc: freebsd-arch@freebsd.org Subject: Re: aio in GENERIC? Message-ID: <CAJ-VmomctCCDf28esRApJXUOeeQ2dZoTR9Rko_9M5a=QOu=xsg@mail.gmail.com> In-Reply-To: <500830D5.6000409@feral.com> References: <3CE55F29-A5B2-44A7-8854-1ED38BAE6F16@FreeBSD.org> <50075072.5050906@gmail.com> <500752CD.9030107@feral.com> <CAJ-Vmo=n=5LAcMS=WPBKNpPB3HZmn2UK9z_gS0R_2bFwHGMEag@mail.gmail.com> <500830D5.6000409@feral.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 19 July 2012 09:07, Matthew Jacob <mj@feral.com> wrote: > On 7/19/2012 8:58 AM, Adrian Chadd wrote: >> >> .. the API supports the avoidance of copyin/copyout? >> >> > okay, I'll buy that, although I'm not certain of the utility in general of > that. > > The reason I asked was that I have rarely, if ever, run into an AIO > implementation that was more trouble than it was worth. It certainly has > caused major grief in FreeBSD when used with CAM target mode. Because the userland code was wrong, or because supporting it in the kernel for CAM target mode is hard? I used posix AIO for Squid servers. It dropped CPU use quite noticably. Adrian
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-VmomctCCDf28esRApJXUOeeQ2dZoTR9Rko_9M5a=QOu=xsg>