Re: [Qemu-devel] write_zeroes/trim on the whole disk

[Alex Bligh]

> On 24 Sep 2016, at 17:52, Alex Bligh <address@hidden> wrote:
> 
> In *your* use-case holes may be desirable. However in the general case, you 
> cannot assume a server supports holes. Optional support for holes isn't even 
> in the mainline spec yet (AFAIR).

You should also be aware that the minimum granularity for writes (currently 1 
byte in the spec, but the blocksize extension allows for larger) may be smaller 
than the minimum granularity for holes, which may depend on the underlying 
filing system.

A hole is in essence merely an optimised way of storing zeroes. And the trim 
operation says 'I'm not worried about this data any more - so zero it if you 
like'.

-- 
Alex Bligh