On Sun, Jun 03, 2018 at 09:12:46AM +0000, Visa Hankala wrote: > On Sun, Jun 03, 2018 at 10:13:47AM +0200, Stefan Sperling wrote: > > On Sun, Jun 03, 2018 at 01:38:47AM -0600, Frederic Cambus wrote: > > > CVSROOT: /cvs > > > Module name: src > > > Changes by: fcam...@cvs.openbsd.org 2018/06/03 01:38:47 > > > > > > Modified files: > > > sys/arch/loongson/conf: RAMDISK > > > > > > Log message: > > > Enable bwfm(4) for USB devices on RAMDISK as well. > > > > > > Requested by visa@. > > > > > > > How was this change tested? > > > > There were reasons why this wasn't enabled in GENERIC and RAMDISK_CD > > on amd64 at the same time. > > > > I don't know if bwfm(4) is ready for ramdisks yet. > > It might be, but I'm not very sure about it. > > E.g. can it gracefully recover from firmware missing upon boot now? > > If the lack of firmware is a problem, then how is that specific > to ramdisks?
In ramdisks, firmware is never present even if rootfs is mounted. Firmawre is only present during upgrades when the installed system's rootfs gets mounted at /mnt. So the driver will always get an error upon the first loading attempt, and will have to retry later. There's a risk of ramdisk-specific bugs where the interface won't recover. That's why I asked what test had been done. I already got a satisfactory answer (fcambus did a full upgrade over bwfm on USB). There was a problem with the driver where no interface would be created if the firmware was not present at the time the device is attached. This has been fixed by patrick since, so all seems good now, for USB devices at least. I don't know if this situation has been tested with bwmfm PCI devices yet *in the ramdisk*. If that works then we should also enable bwfm in bsd.rd for amd64/arm64.
