On 14 August 2013 09:52, Alexander Graf <agraf@xxxxxxx> wrote: > On 13.08.2013, at 14:03, Peter Maydell wrote: >> An obvious thing this machine does not provide is a serial >> port. I would rather just use virtio-console (and we >> should implement the 'emergency console/earlyprintk' bit of >> the virtio spec). > > Are you sure about this? Nope, which is why I flagged it up in the commit message. "No UART" is the idealistic approach, but maybe it won't work... > Not implementing a UART / something more > standardized has been an absolute nightmare on s390. Device names > diverge, so distributions get confused Device names are all over the shop for ARM serial ports anyway: no two boards are the same. (This is a Linux kernel general design flaw -- "first serial port" should be the same /dev/ name regardless of actual hardware/driver implementing it.) That said, "we tried this on s390 and it was a mistake" is a pretty good argument. > and there's this nasty bug > somewhere in virtio-console that makes your input lag when you enter > a lot of data quickly on the port. > > I would really prefer to just define a UART for this machine. It > will make life a lot easier. UARTs are awkward because you have to get into defining clocks for them in the device tree as far as I can see. And you still have the problem that the kernel doesn't know where they are for earlyprintk purposes. -- PMM _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
