Re: [RFC] splice/tee: len=0 fast path after validity check

[Jens Axboe <axboe@kernel.dk>]

On 5/9/20 2:46 AM, Pavel Begunkov wrote:
> When len=0, splice() and tee() return 0 even if specified fds are
> invalid, hiding errors from users. Move len=0 optimisation later after
> basic validity checks.
> 
> before:
> splice(len=0, fd_in=-1, ...) == 0;
> 
> after:
> splice(len=0, fd_in=-1, ...) == -EBADF;

I'm not sure what the purpose of this would be. It probably should have
been done that way from the beginning, but it wasn't.  While there's
very little risk of breaking any applications due to this change, it
also seems like a pointless exercise at this point.

So my suggestion would be to just leave it alone.

-- 
Jens Axboe