On Fri, Dec 11, 2015 at 05:47:15PM -0700, Jason Gunthorpe wrote: > On Fri, Dec 11, 2015 at 07:23:13PM -0500, ira.weiny wrote: > > On Fri, Dec 11, 2015 at 05:00:47PM -0700, Jason Gunthorpe wrote: > > > > > > FWIW, I also hate the sysfs counters that reflect the PMA, these would > > > be much better are free running, wrapping, non-resetting counters > > > unrelated to the PMA. Something that doesn't zero after the SM samples > > > it. Sounds like qib, hif, rdmavt can trivially fix this, and should, IMHO. > > > > To be fair with a 64bit counter these are not going to get reset very often. > > It resets when ever the SM sends a reset packet, so 'whenever'
It's been a while since I have looked at OpenSMs PM but a 64 bit counter does not get reset very often. For OPA with 64bit counters the same is true. > > > Furthermore, I don't think we can change the behavior now. > > Sure we can, the restting is really a bug, to the point that nothing > can actually use the sysfs counters reliably. Well... That would be changing the behavior that everyone is expecting. I know your argument; these are useless and no one is using them. But that is not the point. They were there as a check to see the PMA counters directly. Any change now is breaking the existing ABI. Don't get me wrong. I'm not against keeping running counters for devices. But, I would rather see a more flexible interface. Something like netlink perhaps. And I will again mention that both IB and OPA have PMs which keep running counters. For now Christophs series is a welcome improvement. Ira -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html