copy on write for splice() from file to pipe?

[Jens Axboe]

On 2/10/23 1:36 PM, Linus Torvalds wrote:
> On Fri, Feb 10, 2023 at 12:32 PM Jens Axboe <axboe at kernel.dk> wrote:
>>
>> No, we very much do have that for io_uring zerocopy sends, which was in
>> the bit below you snipped from the reply. It'll tell you when data has
>> been sent out, and when the data has been acked.
> 
> Hmm. splice() itself definitely doesn't have that data - there's no
> "io context" for it.
> 
> There is only the pipe buffers, and they are released when the data
> has been accepted - which is not the same as used (eg the networking
> layer just takes another ref to the page and says "I'm done").
> 
> Maybe adding some io context to the pipe buffer would be possible, but
> it's certainly not obvious how, without changing splice() semantics
> completely.

Right, I'm referencing doing zerocopy data sends with io_uring, using
IORING_OP_SEND_ZC. This isn't from a file, it's from a memory location,
but the important bit here is the split notifications and how you
could wire up a OP_SENDFILE similarly to what Andy described.

-- 
Jens Axboe