SPAM: Re: XFS corruption during power-blackout
Quoted 3 lines · Chris Wedgwood
If caching is enabled I still lose data. Linux does have a concept of write barriers but these are presently not implemented for XFS right now.
I implemented them some time ago for log writes in XFS. Not for fsync though, although fsync usually does a log write afterwards so it should work in practice too. fdatasync might not.
Don't know if the code hasn't bit rotted away and it also was a bit dumb. It was definitely there at some point.
But then a lot of ATA disks and SCSI don't support barriers. Or at least the IDE barrier tests fails on several of my machines.
-Andi
Quoted 2 lines · Andi Kleen
Don't know if the code hasn't bit rotted away and it also was a bit dumb. It was definitely there at some point.
It did rot away sadly but someone (speak up!) was working on a newer version of this I just don't know how much time they've had to work on this recently.
Quoted 2 lines
But then a lot of ATA disks and SCSI don't support barriers. Or at least the IDE barrier tests fails on several of my machines.
IDE has no barriers. I thought the kernel was aware of this and flushed when in those cases? Does that not work?