Re: [Qemu-block] [Qemu-devel] [PATCH 2/2] qemu-img: Add dd seek= option

[Eric Blake]

On 08/15/2018 09:39 PM, Eric Blake wrote:

> > (My full opinion on the matter: dd has a horrible interface.  I don't
> > quite see why we replicated that inside qemu-img.  Also, if you want to
> > use dd, why not use qemu-nbd + Linux nbd device + real dd?)
>
> Because of performance: qemu-nbd + Linux nbd device + real dd is one 
> more layer of data copying (each write() from dd goes to kernel, then is 
> sent to qemu-nbd in userspace as a socket message before being sent back 
> to the kernel to actually write() to the final destination) compared to 
> just doing it all in one process (write() lands in the final destination 
> with no further user space bouncing).  And because the additional steps 
> to set it up are awkward (see my other email where I rant about losing 
> the better part of today to realizing that 'dd ...; qemu-nbd -d 
> /dev/nbd1' loses data if you omit conv=fdatasync).

Oh, and because the kernel NBD module still doesn't know how to do 
sparse writes or expose the location of holes. In the kernel source, see 
drivers/block/nbd.c and weep at how far it lags behind qemu-nbd features 
that at least 'qemu-img convert' can take advantage of, but aren't 
present via /dev/nbd*

--
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org