From mboxrd@z Thu Jan 1 00:00:00 1970 From: paulmck@linux.vnet.ibm.com (Paul E. McKenney) Date: Fri, 13 Feb 2009 06:56:54 -0800 Subject: [ltt-dev] [RFC git tree] Userspace RCU (urcu) for Linux (repost) In-Reply-To: <200902140050.44550.nickpiggin@yahoo.com.au> References: <20090212023308.GA21157@linux.vnet.ibm.com> <20090212215959.GN6759@linux.vnet.ibm.com> <200902140050.44550.nickpiggin@yahoo.com.au> Message-ID: <20090213145653.GA6854@linux.vnet.ibm.com> On Sat, Feb 14, 2009 at 12:50:43AM +1100, Nick Piggin wrote: > On Friday 13 February 2009 08:59:59 Paul E. McKenney wrote: > > On Thu, Feb 12, 2009 at 01:15:08PM -0800, Linus Torvalds wrote: > > > On Thu, 12 Feb 2009, Paul E. McKenney wrote: > > > > In other words, you are arguing for using ACCESS_ONCE() in the loops, > > > > but keeping the old ACCESS_ONCE() definition, and declaring BF hardware > > > > broken? > > > > > > Well, I _also_ argue that if you have a busy loop, you'd better have a > > > cpu_relax() in there somewhere anyway. If you don't, you have a bug. > > > > > > So I think the BF approach is "borderline broken", but I think it should > > > work, if BF just has whatever appropriate cache flush in its cpu_relax. > > > > OK, got it. Keep ACCESS_ONCE() as is, make sure any busy-wait > > loops contain a cpu_relax(). A given busy loop might or might not > > need ACCESS_ONCE(), but that decision is independent of hardware > > considerations. > > > > Ah, and blackfin's cpu_relax() does seem to have migrated from barrier() > > to smp_mb() recently, so sounds good to me!!! > > > Interesting. I don't know if you would say it is not cache coherent. > Does anything in cache coherency definition require timeliness? Only > causality I think. > > However I think "infinite write buffering delay", or requiring "cache > barriers" is insane to teach any generic code about. BF would be free > to optimise arch functions, but for correctness surely it must also > have a periodic interrupt that will expose stores to other CPUs. I have great sympathy for this point of view!!! So why not have the blackfin folks get the appropriate instructions added in the gcc port to their architecture? (Yeah, I know, gcc has no way of knowing which variables are shared and not...) But perhaps we could decorate the affected variable declarations with a macro that expands to some sort of gcc attribute in the blackfin case? Thanx, Paul