Manuel Bouyer <bou...@antioche.eu.org> wrote: > > Few minor points. Does atabus_alloc_drives() need to use M_NOWAIT and > > malloc(9), instead of kmem(9)? > > At first glance I'd say it should always be called from the atabus thread, > so it should be OK
Could you change it then? Generally, KM_NOSLEEP should not be used unless there is a very good reason. It may fail even if there is free memory. > > Also, any reason why you use tsleep(9) > > instead of kpause(9) (in satapmp_subr.c and elsewhere)? We are trying > > to get rid of the former. > > kpause(9) could be used here. But I prefer to keep sys/dev/ata consistent > and convert all the code to kpause at the same time. It could have been > done here, I just didn't think about it. > Next opportunity will be with NCQ support, which I hope to look at soon. > This also would be the time to make ata(4) MP-safe. Cool! NCQ and MP-safe ata(4) would be a major improvement for NetBSD! -- Mindaugas