On Tue, 4 Dec 2012 11:09:36 -0200 Herton Ronaldo Krzesinski <[email protected]> wrote:
> Hi, I got a request that following commit is good for stables. I also > agree that is applicable. Can this be acked and if so queued to all > stables? I verified that at least starting with 3.0 it should be > applicable with minor adjusting, 3.2 onwards should cherry-pick cleanly: > > commit bc78c57388e7f447f58e30d60b1505ddaaaf3a7d > Author: Denis Efremov <[email protected]> > Date: Thu Oct 11 13:08:02 2012 +1100 > > md/linear: rcu_dereference outside read-lock section > Why? That patch doesn't change behaviour, doesn't fix a data-corruption bug, doesn't fix a crash. All it does is: According to the comment in linear_stop function rcu_dereference in linear_start and linear_stop functions occurs under reconfig_mutex. The patch represents this agreement in code and prevents lockdep complaint. i.e. prevents a lockdep complaint. This is important for mainline but irrelevant for -stable. So unless you can explain why stable needs this, I have to say "no - not for stable". NeilBrown
signature.asc
Description: PGP signature
