Re: [RFC bpf-next v4 0/7] Introduce eBPF support for HID devices (new attempt)
From: Alexei Starovoitov <hidden>
Date: 2022-05-13 19:43:14
Also in:
bpf, lkml
From: Alexei Starovoitov <hidden>
Date: 2022-05-13 19:43:14
Also in:
bpf, lkml
On Fri, May 13, 2022 at 10:02 AM Benjamin Tissoires [off-list ref] wrote:
On Thu, May 12, 2022 at 6:23 AM Alexei Starovoitov [off-list ref] wrote:quoted
On Sat, Apr 30, 2022 at 09:12:09AM +0200, Benjamin Tissoires wrote:quoted
Also, I wonder if we should not have some way to namespace kfuncs. Ideally, I would like to prevent the usage of those kfuncs outside of some helpers that I define in HID so I don't have to worry too much about other trace programs fuzzing and segfaulting the kernel.That would be a great feature to have. Other folks expressed the same interest. Just grouping them by prog type is not flexible enough. It feels kfuncs could be scoped by (prog_type, attach_btf_id or attach_hook) pair. What are your thoughts?Scoping by attach_btf_id is very appealing to me (attach_hook less TBH): I have internal functions I do not want normal users to use, and also it would also restrict who can call what in the more general case. However, I don't think I'll put that effort in v5. It is a nice to have feature IMO, but not really required ATM.
Great. Looking forward to it.