On 3/5/10, Jon Masters <jcm@xxxxxxxxxx> wrote: > On Fri, 2010-03-05 at 10:31 +0000, Alan Jenkins wrote: >> On 3/4/10, Ozan Çağlayan <ozan@xxxxxxxxxxxxx> wrote: > >> > http://bugs.gentoo.org/attachment.cgi?id=189127&action=view >> > >> > Shortly I think that m-i-t should track this pulled-in dependency chain >> > and >> > could be able to prune them all upon >> > unloading. According to the manpage of modprobe, -r seems to handle this >> > if >> > the dependencies are not used too. Maybe >> > it's just a missed use-case where the dependencies have refcount >= 0 >> > and >> > not refcount == 0. >> >> Ah. After trying this myself, I see how annoying this could be. >> There's not really any better way to do this without kernel >> modifications, sorry. > > Right. This is a known problem. And it's something I was thinking about > after sending the other mail. I agree that this could do with fixing. As > you say, a lot of this will get better with softdeps, but there's always > the case that a module load triggers a uevent that loads another module > we will now depend upon, or several other possibilities. > > Let's start with softdeps (and look at hinting in kernel modules), then > the next step will be to address the deps we read from the kernel so > that we know who the additional user is in order to unload. > > Jon. Sure. I wasn't saying it was a good idea to modify the kernel for this :). Unloading is only for debugging nowadays. That said, it might be nice to have "modprobe -r a b c" do the same as the gentoo script: repeatedly select modules with refcnt=0 for removal (and then complain about the rest). Then you can do modprobe -r `lsmod | cut -d " " -f 1 | grep snd` which would be easier on the fingers and/or eyes. Alan -- To unsubscribe from this list: send the line "unsubscribe linux-modules" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
