Hi, On 31/10/18 16:18, Linus Torvalds wrote:
On Tue, Oct 30, 2018 at 10:34 PM Al Viro <viro@xxxxxxxxxxxxxxxxxx> wrote:mount API series from David Howells. Last cycle's objections had been of the "I'd do it differently" variety and with no such differently done variants having ever materialized over several cycles...Having just lurked on the discussions about the bugs in here, I don't think this is ready. Maybe they got fixed, but if so, it was recent and it was pretty fundamental. The stated aim of the series is to make the mount API _better_, not worse. And right now it looks like a "two steps back, one theoretical step forwards" kind of better. Linus
The design of the new mount API has been under discussion for some time, both on the mailing lists, and also at LSF/MM too. Al and David (and others) have put a lot of work into getting to the current position, and have specifically requested input from Eric about his concerns over past cycles.
When I look at the discussions I'm seeing two main issues (please correct me if you think I'm wrong about this) which are (a) whether the design is correct and (b) whether there are still bugs in the current patch set.
Which of these are you most concerned about? It seems to me that there is not a lot of point in spending a large amount of time in additional review/testing of the current patch set if the overall design is set to be rejected. If your concerns are only with the robustness/stability of the patch set, then that provides a clear route to resolving the current impasse, at least assuming that Eric is able to enumerate the issues that he has discovered.
It looks like David has already provided a fix for one of the issues which Eric mentioned in his recent email. Eric it would be good if you could confirm that this addresses that particular concern,
Steve.
