Messages in this thread |  | | Date | Tue, 5 May 2026 16:35:52 +0800 | | From | Cheng-Yang Chou <> | | Subject | Re: [PATCH v2 sched_ext/for-7.1] sched_ext: Invalidate dispatch decisions on CPU affinity changes |
| |
Hi Andrea,
On Mon, May 04, 2026 at 11:58:21PM +0200, Andrea Righi wrote: > On Mon, May 04, 2026 at 11:24:57AM -1000, Tejun Heo wrote: > > On Mon, May 04, 2026 at 08:00:50AM +0000, Kuba Piecuch wrote: > > > Random thought: If exposing qseq values to BPF directly is undesirable, then > > > perhaps a less objectionable approach would be to expose them as opaque > > > cookie/token values? Same semantics, but fewer SCX internals leaking to BPF. > > > > Yeah, cookie sounds fine to me and let's clearly note that this is for > > schedulers that don't implement properly synchronized dequeue. > > I also like the idea of exposing cookies and extending scx_bpf_dispatch_cancel() > to support direct dispatches, it it doesn't introduce too much complexity (it > shouldn't be too painful). > > Cheng-Yang, just to make sure we're not doing the work twice, are you planning > to work on this?
If you are already working on this, I can help with testing after your patches! Otherwise, I'm happy to help.
FYI, I'm already working on deferring ext_server bandwidth to SCX enable and releasing it when disabled, but it still needs more testing!
-- Cheers, Cheng-Yang
|  |