Matthew Burgess wrote:
As you know, I'm still a little unsure of how the udev and hotplug things all hang together (or not as the case may be!). Recent developments in udev suggest that the hotplug package might not be required anymore as udev can handle hotplug events itself.
Right. But it won't handle it in exactly the same way as hotplug, and it will break everything in BLFS that depends upon usbmaps (e.g. gPhoto2). Upstream recommendation: wait until the equivalent of obsolete /proc/bus/usb/xxx/yyy pseudofiles is accessible in /dev (2.6.14? anyway, that's already in -mm), then replace these usbmaps with udev rules. The "10000 udev rules needed to correctly chmod all scanners and cameras" problem is still not solved.
However, RELEASE-NOTES says: "For full support, the broken input-subsytem needs to be fixed, not to bypass the driver core." This looks as if it's going to get into 2.6.15 as there were a series of patches (starting at http://www.ussg.iu.edu/hypermail/linux/kernel/0509.1/2566.html) which I think do exactly this.
That's really a non-issue if input drivers are not modules.
So, does anyone with more of a clue about this stuff no for sure what the current state of play is? If we removed hotplug now, would that simply force folks to not have any input drivers as modules, or are the effects of the broken input-subsystem more subtle? What advice can we give someone who wants to use a modular kernel regarding what their udev rules and modprobe.conf should look like?
Exactly the same recommendations as now. Use "install" lines to pull optional dependencies (like snd_pcm -> snd_pcm_oss). Don't use aliases for on-demand module loading because that won't work. Put everything not detectable by the equivalent of the present hotplug script into /etc/sysconfig/modules (should be only non-hardware devices such as ppp or loop, plus ISA cards).
-- Alexander E. Patrakov -- http://linuxfromscratch.org/mailman/listinfo/lfs-dev FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
