On Sun, Oct 22, 2017 at 7:47 PM, Chris Mi <chr...@mellanox.com> wrote: > > It seems it is not easy to discard call_rcu(). I'm afraid even if we have a > final solution > without call_rcu(), it is not mature at the beginning as well. I mean we also > need time
Why do you believe it is not easy? RTNL lock is already there, list_splice_init_rcu() is there too. I can naturally divide my patches for each module so that they are much easier to backport than yours. > to fix the possible bugs of the new design. And maybe to destroy the filters > in parallel > is the right direction. If this bug is the last bug brought by call_rcu(), > then changing it > may not be a good idea. Again, you have to prove this is the last bug, I seriously doubt it is. > > Patch 1 is straightforward to use atomic. Patch 2 is to convert the list to > array. Both are big in size. > I think there is no harm to the new design. Patch 3 and 4 are useful test > case. It definitely doesn't harm, but why do we waste time on it when we know there is a better way? It is clearly not easy for backport either, in fact it is harder w.r.t. size. > We also need it with new design to make sure there is no regression. > Are you saying I can't trust your test cases? ;) > So I think my patch set should not be held so long time. I think your patches should be dropped except the last two, I will take the last two for you. Thanks!
