From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jim Paris Subject: Re: Warnings on shutdown (Debian Etch and Kernel 2.6.22.9) Date: Thu, 25 Oct 2007 15:45:51 -0400 Message-ID: <20071025194551.GA8566@jim.sh> References: <471DC728.60705@mandic.com.br> <471DC8F1.9090501@gmail.com> <471DCA74.503@mandic.com.br> <471DCB33.5070108@gmail.com> <471DD288.7010902@mandic.com.br> <471DE866.7040908@gmail.com> <20071023234732.GA23598@jim.sh> <471F493C.8050804@mandic.com.br> <471F4B11.9030206@gmail.com> <47208144.9080602@mandic.com.br> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Received: from NEUROSIS.MIT.EDU ([18.95.3.133]:33230 "EHLO neurosis.jim.sh" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753534AbXJYTqZ (ORCPT ); Thu, 25 Oct 2007 15:46:25 -0400 Content-Disposition: inline In-Reply-To: <47208144.9080602@mandic.com.br> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: "Renato S. Yamane" Cc: Tejun Heo , Jeff Garzik , debian-devel@lists.debian.org, linux-ide@vger.kernel.org, miquels@cistron.nl Renato S. Yamane wrote: > Tejun Heo wrote: > >Renato S. Yamane wrote: > >>Jim Paris wrote: > >>>For Debian, I don't think there are any complete solutions yet. > >>>The related bug is =20 > >>>http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=3D426224 > >> > >>I think that we can fix this if someone can answer this: > >> > >>$ man halt > >>"-h flag: This is important for IDE drives, since the kernel doesn=E2= =80=99t=20 > >>flush the write cache itself before power-off.." > >> > >>How can I know if kernel write cache itself? :-) > > > >Kernel has been flushing write cache for many many years now. Man p= age=20 > >needs update. Also, flushing cache from halt is okay. What's hurti= ng=20 > >is spinning down the drive from halt. >=20 > So, I think that halt script need know: >=20 > - How many and who is IDE drives and SCSI/SATA drives. > - Devices with manage_start_stop as "0" or without manage_start_stop=20 > need "halt -h". > - Devices with manage_start_stop as "1" need only "halt". >=20 > This is correct? The real changes need to be made in the shutdown utility, as described here: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=3D426224#10 -jim