On Fri, 02 Jun 2006, martin f krafft wrote: > also sprach Henrique de Moraes Holschuh <[EMAIL PROTECTED]> [2006.06.02.1534 > +0200]: > > Yes, you lose. Don't use XFS for /, or /boot for that matter. It > > is not safe. > > But yes it is. In fact, I consider it the safest around. But > generally only if you have a battery-backed power-supply, which > curiously doesn't apply to laptops the way I mean it.
No, it is not. XFS has bad data-ordering-flush semanthics for resilience, they are superb for speed and performance, NOT for data safety. It won't guarantee data safety unless you fsync(), umount or mount -o ro,remount... and it does *not* guarantee data safety on global sync(), which is what this bug is about. Also, XFS does not tolerate use of its three first sectors for MBR / bootblocks, which is not a defect or anything, but it *is* something you must remember or you will corrupt its first superblock. > Since hibernate only really applies to laptops and desktops, and > they aren't really XFS candidates, I guess the bug is a wontfix for > now. :( If you *do* take the time to make sure sync() is NOT flushing everything XFS to disk, *please* file a bug on kernel.org, otherwise it will never be fixed where it really matters. I still think an hibernate scriptlet to mount partitions readonly (instead of just umounting them) is a valid, useful wish. And it would fix the issue with /boot as XFS (but not of / as XFS). -- "One disk to rule them all, One disk to find them. One disk to bring them all and in the darkness grind them. In the Land of Redmond where the shadows lie." -- The Silicon Valley Tarot Henrique Holschuh -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
