On Mon, 2006-06-03 at 15:01 +0100, Patrick McHardy wrote: > > xfrm_user can be a module, so an ifdef won't help.
good point. > The rcu_dereference > doesn't cost anything except on alpha. We could reduce the overhead a > bit more by avoiding the rcu_read_lock/rcu_read_unlock, the code is > AFAICT only called from the xfrm output functions, which already have > preemption disabled in xfrm[46]_output. But I'm always a bit reluctant > to do things like that, it makes it harder to understand and is sure to > be forgotten when the surrounding code changes. My only concern is performance. If it is negligible then no issues. In regards to the rcu_read_[un]lock perhaps an alternative is you comment out the rcu_read_[un]lock and leave it in the code with some explanation text to serve as a reminder for whoever makes future changes? Again, if it is negligible then this is not needed (I do plan to run some performance tests and could have used the opportunity to check differences but that wont happen for another 2 weeks) cheers, jamal - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
