On 1/21/22 17:04, Vladimir Sementsov-Ogievskiy wrote:
The split was proposed in previous versions, but Vladimir did not
really like it and suggested to send it as a separate series:
I didn't really like it as it seemed unusual and unobvious to me. But if
we already accepted similar split for generic block layer, no way for me
to resist :) And if we follow new logic of generic block layer in jobs,
it's not "unusual" any more.
Either way I think it's okay to have it as a follow-up. The explicit
naming in the API is a bit verbose but definitely clearer, so it's okay
to order different than the graph/IO split. In that case we weren't
even sure, until you went through all the testcase failures, that a
_locked or rather "_drained" API was possible.
Paolo