From mboxrd@z Thu Jan 1 00:00:00 1970 From: Paul van der Vlis Subject: Re: Software raid, booting and bios Date: Fri, 20 May 2011 17:53:04 +0200 Message-ID: References: <9067914580344941270@unknownmsgid> <20110520145623.749b4781@natsu> <4DD65A66.5030904@turmel.org> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <4DD65A66.5030904@turmel.org> Sender: linux-raid-owner@vger.kernel.org To: linux-raid@vger.kernel.org List-Id: linux-raid.ids Op 20-05-11 14:11, Phil Turmel schreef: > On 05/20/2011 05:33 AM, Paul van der Vlis wrote: >> >> The problem is about detected disks with a defect in the MBR. > > This is a crucial point. A BIOS that supports multiple drives in the > boot order should skip to the next if the MBR cannot be read. But > the BIOS loses control once the MBR code is executed. If an error is > encountered in later sectors of the bootloader, there's no way to > switch to the next drive. This is also true when the BIOS only > supports dissimilar devices in the boot order. I think you are right. > If I had to minimize the chance of this ever biting me, I'd use a CF > <==> IDE adapter with a DMA capable CF card, and set it up as my boot > device. And I wouldn't use it for anything but boot. A quick google > turned up this: > > http://www.addonics.com/products/flash_memory_reader/adidecf.asp In the servers I use, there is normally no place for such a cardreader. But a low profile PCI card with a CF card on it could do it. > (Just to show what's out there.) The embedded boards I use > occasionally have the equivalent of this soldered to their > motherboards. > > The best DMA capable CF cards are usually found in markets that cater > to industrial designers or to professional photographers. Do you think the risk of a problem with a CF card (or something like that) is much lower then the risk of a problem with a harddisk? And what about booting from an USB stick? With regards, Paul van der Vlis. -- http://www.vandervlis.nl