Re: [Qemu-devel] [PATCH] sheepdog: implement direct write semantics

[Stefan Hajnoczi]

On Tue, Jan 08, 2013 at 01:42:35PM +0800, Liu Yuan wrote:
> On 01/07/2013 09:23 PM, Kevin Wolf wrote:
> > No, and in theory they shouldn't have to care.
> > 
> > Why do you want to handle writethrough semantics in the block driver
> > rather than letting qemu care for sending the right flushes?
> 
> Because Sheepdog itself provide the client side cache implementation and
> we need means to get the cache hint from users when starting up the VM.
> This cache support:
>  writethrough: provide a read cache for this VM, which is always
> consistent with cluster data
>  writeback: provide a writeback cache for this VM, which only flush the
> dirty bits when VM send 'flush' request.
>  directio: disable cache completely for this VM.
> 
> Old QEMU (before v1.2) allows block drivers to get the cache flags, so
> Sheepdog can interpret these flags on its own to choose the right cache
> mode for each VM. doesn't QEMU need keep backward compatibility?

This sounds fine - it fits together with QEMU like this:

BDRV_NO_CACHE means sheepdog directio.  We're disabling the client-side
sheepdog cache.

Otherwise use sheepdog writeback and let QEMU block.c decide when to
flush.  Never use sheepdog writethrough because it's redundant here.

Stefan