On Fri, Oct 13, 2017 at 10:27:50AM +0200, Peter Zijlstra wrote: > On Fri, Oct 13, 2017 at 04:56:33PM +0900, Byungchul Park wrote: > > On Thu, Oct 12, 2017 at 05:56:35PM +0200, Peter Zijlstra wrote: > > > On Thu, Oct 12, 2017 at 08:38:17AM -0700, Tejun Heo wrote: > > > > > > > > As long as we have the same level of protection, simpler code is of > > > > course preferable. That said, I haven't followed the discussion > > > > closely and don't want to apply it without Peter's ack. Peter? > > > > > > I'm really tied up atm; and feel we should be addressing the false > > > positives generated by the current code before we start doing new stuff > > > on top. > > > > We can never avoid adding false dependencies as long as we use > > acquisitions in that way the workqueue code does, even though you > > successfully replace write acquisitions with recursive-read ones after > > making them work, as you know. > > Not the point; they still need to get annotated away. The block layer > and xfs are now fairly consistently triggering lockdep splats, that > needs to get sorted.
I'm not sure if I can help, since I'm not familiar with that sub-systems, but I want to do something for them if I can. Could you share the issues and test cases? Or I suggest that don't wait for recursive-read work to complete but use 'might' thing to avoid the false annotations. I don't say it to rush you to take my patches, but I do beacuse it does exactly what we need in that case. I believe it would be helpful.
