On Thu, Sep 16, 2021 at 7:06 AM Jakub Kicinski <kuba@xxxxxxxxxx> wrote: > > On Wed, 15 Sep 2021 20:21:04 -0700 Alexei Starovoitov wrote: > > +In HW > > +----- > > + > > +The HW can choose to execute eBPF instruction natively and provide eBPF runtime > > +in HW or via the use of implementing firmware with a proprietary license. > > That seems like a step back, nfp parts are all BSD licensed: Yeah. netronome is a great example of how firmware should be developed. > > +Packaging BPF programs with user space applications > > +==================================================== > > + > > +Generally, proprietary-licensed applications and GPL licensed BPF programs > > +written for the Linux kernel in the same package can co-exist because they are > > +separate executable processes. This applies to both cBPF and eBPF programs. > > Interesting. BTW is there a definition of what "executable process" is? That's how lawyers put it. BPF in many ways is unique, so traditional computer science words and meanings don't apply 100%. Like, bpf programs are analogous to kernel modules, but they're very different at the same time as well. The analogies are used to explain things. Is bpf program an "executable process" on its own? Hard to say. Is there a task_struct allocated for each bpf prog? Of course, not. But it is certainly not executing in ring 3. In the future we might have kthread or even user thread completely occupied by bpf progs.
