> It seems like a dead and and there won't be a consensus on how to implement > this. > Without disrespecting Jason's opinion I still would like to see this patch > get in. > What's the convention in such cases?
In general when the maintainer (ie me) is not convinced about a patch and there is not a strong consensus from others then we don't merge the patch. And in this particular case I find myself agreeing more and more with Jason. The big issue seems to be your requirement that this debugging work with "cat" with no other tools needed; however sticking stuff in debugfs means also that "mount" as root is needed. And to be honest, putting hard-to-extend and hard-to-parse text blobs in debugfs does seem much less useful than a netlink interface (which is much easier to use programmatically and allows for extension), even if netlink requires someone to write a (simple) userspace tool to dump information. OFED gives you a short-term way to make sure that tool is available to your users, and longer-term integrating with existing tools will be easier if you use netlink. Really, is being able to dump with "cat" so important? Even if it is, I would still prefer to see the infrastructure for doing this properly through netlink go in first, and add the debugfs support on top of that. As an aside, I think adding ftrace events support to the cma might be useful for debugging as well. - R. _______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
