First patch will (try to) remove lost busoff conditions by not reading the status register wihtout pending status interrupt. This has ran for several weeks, and not produced any lost busoff (I'm not sure it would have issued a true positive in this time). I run this on a beaglebone-alike board. Next patch will tear c_can bottom halve isr out of napi so I can better control it's RT priority without affecting other napi handlers in the same softirq. Last patch will combine top and bottom halve isr in case of force irq threading like on an RT kernel. Having the irq split across 2 threads does not seem wise. Changes since v1: I refactored the 1/3 to avoid a merge conflict with patches of Jeroen Hofstee. Kind regards, Kurt
