On Thu, May 11, 2017 at 10:45:27PM +0200, Martin Wilck wrote: > Except that it can cause userspace to miss events, as I tried to point > out in my reply to Mike. It should never miss new events. Because the calls are separate, occasionally it might trigger when it didn't need to, but userspace will see nothing new happened and just go back into poll. Combining the two calls would avoid this, but the authors decided on balance it was unnecessary. I've already requested a libdevmapper/dmsetup implementation as a matter of urgency so that everyone has easy command-line access to the new features and example code and can test them - normally patches like these shouldn't go upstream until a supported userspace implementation is available. Once we have this, questions like the ones you pose can be checked practically. Alasdair -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
