> > On Wed, Jan 02, 2019 at 10:54:15AM -0700, Jason Gunthorpe wrote: > > On Thu, Dec 27, 2018 at 09:05:50PM +0000, Weiny, Ira wrote: > > > > > Regardless, and more importantly, this is a totally different thing. > > > eBPF sample _code_ is a long way from a tool which is dependent on > > > this "ABI". > > > > There has been some general debate in the kernel community if > > tracepoints are ABI or not ABI (like debugfs) and I don't know where > > it ended up.. > > > > Core tracepoints are probably more likely to be ABI ones, I guess.. I agree they will be "more likely" to be ABI but as we are starting out here we need to leave some flexibility... > > Tracepoints are valuable debug tool and declaring them as ABI will create > strange situation where refactoring/improvements can be blocked simply > because of those debug aids. > > ABI thing will effectively kill all additions of tracepoints to RDMA/core. I agree. Right now these tracepoints were good enough for the OPA debugging I have done over the last few years. But they have not been generally used because building a kernel with them was too much trouble for many users. It will not be until we get some feed back and usage when we will know if the amount of information in the trace is too much or too little. Ira > > Thanks > > > > > Jason
