[dm-devel] [PATCH] reworked dm-switch target

[Mikulas Patocka]

On Tue, 11 Sep 2012, Jim Ramsay wrote:

> 
> On Sat, Sep 08, 2012 at 12:35:22PM -0400, Mikulas Patocka wrote:
> > Your changes are ok, it could be included in the kernel.
> > 
> > Regarding flushes ... could you please explain how write-back caching 
> > works in your design?
> 
> We are not actually using write-back caching in our design at all, but
> assumed that for a general-purpose DM device like dm-switch, some
> underlying storage may rely on REQ_FLUSH to ensure proper data
> synchronization, so the switch should do the right thing when required.
> The code is modeled on dm-stripe.

So you don't need to process REQ_FLUSH at all. REQ_FLUSH doesn't impose 
any request synchronization, it only flushes the hardware cache - and 
there is no cache. I would remove REQ_FLUSH support if it has no use.

dm-switch is used directly on host adapters (or maybe dm-mpath that is 
connected on host adapters), there is no intermediate layer that would 
need REQ_FLUSH.

Mikulas

> > Is there some write-back cache in the host adapter?
> 
> No, there is not at this time.
> 
> > Is there some write-back cache in the individual storage nodes? (i.e. does 
> > the node signal end of I/O before it finishes writing data to the disk) 
> 
> No, there is not at this time.  The iSCSI-backed SD devices we are
> ultimately using do not advertise any write-back cache ('sdparm
> --get=WCE' returns 0).
> 
> > Another thing - please resend your code with "Signed-off-by". Read the 
> > meaning of "Signed-off-by" in Documentation/SubmittingPatches, agree to 
> > the terms and append "Signed-off-by: Jim Ramsay <jim_ramsay at dell.com>" to 
> > the code. It is a legal requirement, so that you certify that the code is 
> > under the open source license and that you have the right to distribute 
> > the code.
> 
> Will do shortly.  Thanks!
> 
> -- 
> Jim Ramsay
>