On Thu, Nov 21, 2024 at 7:33 AM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote: > > On Thu, Nov 21, 2024 at 06:35:39AM -0800, Vadim Fedorenko wrote: > > On 21/11/2024 03:32, Peter Zijlstra wrote: > > > On Wed, Nov 20, 2024 at 04:08:11PM -0800, Vadim Fedorenko wrote: > > > > New kfunc to return ARCH-specific timecounter. For x86 BPF JIT converts > > > > it into rdtsc ordered call. Other architectures will get JIT > > > > implementation too if supported. The fallback is to > > > > __arch_get_hw_counter(). > > > > > > Still not a single word as to *WHY* and what you're going to do with > > > those values. > > > > > > NAK > > > > Did you have a chance to read cover letter? > > Cover letter is disposable and not retained when applying patches, as > such I rarely read it. It's not disposable for BPF trees. We preserve them as part of the merge commit for the patch set, e.g. [0]. Both bpf and netdev maintainers use a set of scripts to apply patches (pw-apply, specifically), that does all that automatically. Vadim, Please do another careful pass over commit messages and cover letter. I'd suggest moving the version history into cover letter (see other multi-version cover letter for an example). You can use an example from your BPF selftests as an intended use case (measuring the duration of some BPF piece of logic), and I'd also mention that this is useful to measure the duration of two related BPF events. E.g., uprobe entry and exit, of kprobe entry/exit. kprobe.session and uprobe.session programs are especially well suited for that, as they allow to capture initial timestamp, store it in session cookie, then retrieve it in return probe and calculate the difference. Please also update all the "cycles" references to "time counter", stuff like that. pw-bot: cr [0] https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/commit/?h=bpf-next-6.13&id=379d5ee624eda6a897d9e1f7f88c68ea482bd5fa [1] https://git.kernel.org/pub/scm/linux/kernel/git/dborkman/pw.git/
