On Sun, Jun 10, 2001 at 05:10:53PM +0200, Michael Schmitz wrote: > > > > On my TiBook, BenH' 2.4.5-pre kernel reprobes all ADB devices after > > wakeup. This, I think, is not new; what's changed is the fact that the > > rescan is not done inline in the wakeup code path, but rather a kernel > > thread is started that does the probe. Wakeup is faster, since we don't > > block on the ADB probe. > > But we run pmud before the device wakeup has done. This is extremely bad, > opening /dev/adb should return -EAGAIN while the bus state isn't sane, as > after sleep (there isn't anything user space related to care about before > ADB is initialized on boot).
Er, quickly reading pmud.c, it _looks_ like we open the device (/dev/adb) once and after that always assume it's open/sane. Same w/ /dev/pmu. It also looks like currently we use some 'magic' to do adb stuff, instead of trying to do this via ioctls or something like that.. -- Tom Rini (TR1265) http://gate.crashing.org/~trini/
