Re: [PATCH 1/1] cp: leverage copy_file_range syscall for copy

[Pádraig Brady]

On 18/07/19 18:53, Olga Kornievskaia wrote:

> Netapp doesn't currently have an NFSv4.2 spec implementation of copy
> offload not to say it won't have one in the future. This has been pure
> Linux to linux NFSv4.2 copy offload support that I'm putting in.

Thanks for that info.

> Also thank you for your feedback and the pointer. Is the fact that
> there is no consistent interface (i.e. kernel version dependency and
> libc (non)support dependencies) means we shouldn't move forward in
> including the feature?
> 
> Is the path forward to take Goldwyn's patch since it supports sparse
> files and then add checking for presence of libc support? I just don't
> want to go down a path that's not seen as the right path by the
> coreutils maintainers.

copy offload is such a useful feature we will be incorporating it.
It's just a matter of when and how to best do it.
cp is such an important program we have to be ultra considerate
of how robust it is in all cases.

Consider the kernel implementation issues discussed at 
https://lwn.net/Articles/774114/

Now things are solidifying in this area finally,
so we'll need to look into how to most robustly enable this
given all the mentioned implementation combinations.

cheers,
Pádraig