On Tue, Aug 9, 2022 at 6:15 PM David Vernet <void@xxxxxxxxxxxxx> wrote: > > Hi Hao, > > On Mon, Aug 08, 2022 at 11:57:53AM -0700, Hao Luo wrote: > > > Note that one thing that is not included in this patch-set is the ability > > > to kick the kernel from user-space to have it drain messages. The selftests > > > included in this patch-set currently just use progs with syscall hooks to > > > "kick" the kernel and have it drain samples from a user-producer > > > ringbuffer, but being able to kick the kernel using some other mechanism > > > that doesn't rely on such hooks would be very useful as well. I'm planning > > > on adding this in a future patch-set. > > > > > > > This could be done using iters. Basically, you can perform draining in > > bpf_iter programs and export iter links as bpffs files. Then to kick > > the kernel, you simply just read() the file. > > Thanks for pointing this out. I agree that iters could be used this way to > kick the kernel, and perhaps that would be a sufficient solution. My > thinking, however, was that it would be useful to provide some APIs that > are a bit more ergonomic, and specifically meant to enable kicking > arbitrary "pre-attached" callbacks in a BPF prog, possibly along with some > payload from userspace. David, very sorry about the late reply. Thank you for sharing your thoughts. I am looking at your v2 and understand you need a way to trigger the kernel to consume samples in the ringbuf, which seems a reasonable motivation to me. > > Iters allow userspace to kick the kernel, but IMO they're meant to enable > data extraction from the kernel, and dumping kernel data into user-space. Not necessarily extracting data and dumping data. It could be used to do operations on a set of objects, the operation could be notification. Iterating and notifying are orthogonal IMHO. > What I'm proposing is a more generalizable way of driving logic in the > kernel from user-space. > Does that make sense? Looking forward to hearing your thoughts. Yes, sort of. I see the difference between iter and the proposed interface. But I am not clear about the motivation of a new APis for kicking callbacks from userspace. I guess maybe it will become clear, when you publish a concerte RFC of that interface and integrates with your userspace publisher. Hao
