Re: [PATCH, RFC] virtio_blk: add cache flush command

[Christoph Hellwig <hch@lst.de>]

On Mon, May 11, 2009 at 12:47:58PM -0500, Anthony Liguori wrote:
> But how do we define the data integrity guarantees to the user of 
> cache=writeback+fsync?  It seems to require a rather detailed knowledge 
> of Linux's use of T_FLUSH operations.

It does work the same as for disks with writeback caches.  If a barrier
request completes the filesystem can assume all previous I/O on this
disk has finished before the barrier request, and the barrier request
has finished after these without any later request being finished before
it.  If a cache flush is issued the cache as of that point in time is
flushed (used for fsync).

> Right now, it's fairly easy to understand.  cache=none and 
> cache=writethrough guarantee that all write operations that the guest 
> thinks have completed are completed.  cache=writeback provides no such 
> guarantee.

As said above with barriers it indeed does.

> cache=writeback+fsync would guarantee that only operations that include 
> a T_FLUSH are present on disk which currently includes fsyncs but does 
> not include O_DIRECT writes.  I guess whether O_SYNC does a T_FLUSH also 
> has to be determined.

O_SYNC and O_DIRECT do it at least on XFS due to updating metadata after
the write.  I can't vouch for implementation details on other
filesystems.