Date: Thu, 4 Aug 2005 10:58:38 +0200 From: Maxim Sobolev <sobomax@portaone.com> To: Andrey Chernov <ache@freebsd.org>, Dan Nelson <dnelson@allantgroup.com>, "current@freebsd.org" <current@freebsd.org> Subject: Re: Sub-optimal libc's read-ahead buffering behaviour Message-ID: <20050804085838.GA20122@www.portaone.com> In-Reply-To: <20050804060830.GB21228@nagual.pp.ru> References: <42F0CCD5.9090200@portaone.com> <20050803150117.GD93405@dan.emsphone.com> <42F0E9B2.9080208@portaone.com> <20050804060251.GA21228@nagual.pp.ru> <20050804060830.GB21228@nagual.pp.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Aug 04, 2005 at 10:08:30AM +0400, Andrey Chernov wrote: > On Thu, Aug 04, 2005 at 10:02:51AM +0400, Andrey Chernov wrote: > > No surprize. Single read can be consider as sort of "atomic" operation. > > But between read & seek character device itself can move its pointer in > > anyplace. > > No only move the pinter, but change contents too. Consider reading > /dev/mem Actually this is good example which supports my point, since with the current behaviour when doing 2 subsequent reads each of them less than 4K in size you will get either outdated data from the second read or even worse - mix of the old data and the new data when second read crosses buffer boundary! I think that any buffering should be disabled for reading from chardev no matter how "historic" our current behaviour is. -Maxim
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050804085838.GA20122>