Re: [RFC PATCH 3/4] io_uring/splice: support splice from ->splice_read to ->splice_read

[Ming Lei]

On Mon, Nov 07, 2022 at 11:45:06PM -0800, Christoph Hellwig wrote:
> On Thu, Nov 03, 2022 at 04:50:03PM +0800, Ming Lei wrote:
> > The 1st ->splice_read produces buffer to the pipe of
> > current->splice_pipe, and the 2nd ->splice_read consumes the buffer
> > in this pipe.
> 
> This looks really ugly.  I think you want Linus and Al to look over
> it at very least.

OK, I will Cc Linus and Al in V2.

It is just another case of pipe's producer/consumer model, IMO.

> 
> Also, what is going to happen if your ->splice_read instance does not
> support the flag to magically do something entirely different?

If the ->splice_read() instance doesn't support this feature, then the new
added pipe flag won't be set, this API will return -EINVAL.



thanks, 
Ming