From mboxrd@z Thu Jan 1 00:00:00 1970 From: Chris Mason Subject: Re: How to know whether disks "handle flush requests correctly" Date: Fri, 06 May 2011 09:48:32 -0400 Message-ID: <1304688928-sup-8382@think> References: <4DC3F33F.80309@redhat.com> Content-Type: text/plain; charset=UTF-8 Cc: Paul Schroeder , linux-btrfs To: Josef Bacik Return-path: In-reply-to: <4DC3F33F.80309@redhat.com> List-ID: Excerpts from Josef Bacik's message of 2011-05-06 09:10:23 -0400: > On 05/06/2011 05:13 AM, Paul Schroeder wrote: > > The btrfs wiki Main Page warns that "it is currently possible to corrupt > > a filesystem irrecoverably if your machine crashes or loses power on disks > > that don't handle flush requests correctly." > > > > How do you know if this applies to your drives? Is there a way to test it, > > or a model list, or are newer SATA drives (magnetic, not SSDs) always ok? > > Does it depend on the controller? (I have a SiI 3114, latest BIOS.) > > > > I would also be using btrfs on top of dm-crypt (with the latest release > > kernel). Some kernel versions ago, the message that write barriers aren't > > supported disappeared; can I assume the device mapper / dm-crypt is not a > > problem with regards to flushing? > > > > Yeah if you don't see those messages you can be fairly certain you are > ok. Thanks, The easiest way to tell for sure is to do two tests: dd if=/dev/zero of=/mnt/foo bs=4K count=10000 dd if=/dev/zero of=/mnt/foo bs=4K count=10000 oflag=sync Run this with the filesystem mounted normally and again with the filesystem mounted -o nobarrier. -o nobarrier should be dramatically and hugely faster, almost like we're not writing to the disk at all. -chris