Hi Marc, On Mon, Apr 5, 2021 at 9:41 PM Belisko Marek <marek.belisko@xxxxxxxxx> wrote: > > Hi Marc, > > On Mon, Apr 5, 2021 at 7:31 PM Belisko Marek <marek.belisko@xxxxxxxxx> wrote: > > > > On Fri, Apr 2, 2021 at 9:28 PM Marc Kleine-Budde <mkl@xxxxxxxxxxxxxx> wrote: > > > > > > On 01.04.2021 11:04:25, Belisko Marek wrote: > > > > Is there anything I can tune to have it read faster? Thanks. > > > > > > Indeed, with current net-next/master, you put the NAPI on a per > > > networking device from softirq to threaded mode. > > > > > > | echo 1 | sudo tee /sys/class/net/can0/threaded > > > > > > Then you a separate NAPI thread per networking device that can be tuned > > > with chrt, etc... > > OK I've tse threaded mode for NAPI and for napi:can0 thread I set chrt > > -p 99 <pid> but unfortunately there are still buffer overruns which is > > kind of strange. > > > > For rx-offload branch I was not able yet to run it on my platform > > (have panic during boot which I need to resolve first). > OK I resolved boot issues on my platform and tested using rx-offload > branch but again the issue is present. > It's enough to generate can frames using candeg can0 -g0 and I get > after some time error/overrun errors. > I'm trying now to bump rx-offload patches on top of linux-can-next > testing branch and test it. Was bit fast with response. With rx-offload feature receiving can frames with high speed (small gaps) is working fine as expected. So what are the next steps? I can port rx-offload for c_can to actual master and verify and post patches for review? > > Thanks and BR, marek
