Hi, On Mon, May 13, 2019 at 9:47 AM, Mark Brown <broonie@xxxxxxxxxx> wrote: > On Mon, May 13, 2019 at 09:03:28AM -0700, Doug Anderson wrote: > > On Mon, May 13, 2019 at 9:02 AM Mark Brown <broonie@xxxxxxxxxx> wrote: > > > > I'm not saying the other changes aren't helping, I'm saying that it's > > > not clear that this revert is improving things. > > > If I add a call to force the pumping to happen on the SPI thread then > > the commit I'm reverting here is useless though, isn't it? > > Well, I'm not convinced that that change is ideal anyway and it does > leave you vulnerable to further changes in the SPI core pushing things > out to calling context which feels like it isn't going to be helping > robustness. OK. Here's my plan: in v2 I've still included this revert and you can see how things look. If you hate it as much as you think you will then let me know and I'll send a v3 that avoids to forcing and re-adds the realtime thread to cros_ec. One note just so you're aware: For my particular device I'm not nearly as concerned with latency / throughput as I am concerned with transfers not getting interrupted once started. I've added this explicitly in the commit message now, too. :-) -Doug _______________________________________________ Linux-rockchip mailing list Linux-rockchip@xxxxxxxxxxxxxxxxxxx http://lists.infradead.org/mailman/listinfo/linux-rockchip
