From: Denis Vlasenko Subject: Re: [RFD] What error should FS return when I/O failure occurs? Date: Wed, 18 May 2005 09:54:32 +0300 Message-ID: <200505180954.32470.vda@ilport.com.ua> References: Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Cc: Elladan , fs@ercist.iscas.ac.cn, linux-fsdevel@vger.kernel.org, Kenichi Okuyama , Valdis.Kletnieks@vt.edu Return-path: Received: from 167.imtp.Ilyichevsk.Odessa.UA ([195.66.192.167]:26093 "HELO port.imtp.ilyichevsk.odessa.ua") by vger.kernel.org with SMTP id S262105AbVERGz2 (ORCPT ); Wed, 18 May 2005 02:55:28 -0400 To: Bryan Henderson In-Reply-To: Content-Disposition: inline Sender: linux-fsdevel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Tuesday 17 May 2005 21:10, Bryan Henderson wrote: > >A new, less aggressive sync option which means "okay to cache writes, > >but start writeback at once and write out all dirty data for this device" > >would bring the best of both worlds. > > You mean "optimum compromise." It's the best of neither world. I mean "it will write data without delay but will not trash your flash and will not slow down your apps". Isn't it bad that in "some other OS" I can copy data to the floppy, _floppy start writing at once_, I watch diode and eject floppy when diode is off. Data is coherent. fs is clean. * In Linux, ext2 fs is not clean if I remove floppy without unmount, no matter how many hours it was there with all dirty data committed. (With automounter, it will be clean when fs gets auto unmounted. Sadly, fs may be busy simply because I forgot a Midhight Commander on a xterm with current directory inside mounted fs, thus we return to square one. ext2 will end up dirty) * In Linux, there is a delay before floppy start writing (if !O_SYNC). Or else (if O_SYNC), copying is slow (also not forget recent USB flash trashing issues). I know why it is so and that fixing it is not completely trivial. But I'm not going to pretent that this issue does not exist. -- vda