On 01/12/2020 13:45, Christoph Hellwig wrote:

On Tue, Dec 01, 2020 at 01:36:22PM +0000, Pavel Begunkov wrote:

quoted

Yeah, that's the idea, but also wanted to verify that callers don't
free it while in use, or if that's not the case to make it conditional
by adding a flag in iov_iter.

Can anybody vow right off the bat that all callers behave well?

Yes, this will need a careful audit, I'm not too sure offhand.  For the
io_uring case which is sortof the fast path the caller won't free them
unless we allow the buffer unregistration to race with I/O.

For registered bufs io_uring waits for such requests to complete first,
so it's fine.

-- 
Pavel Begunkov