linux_base and devfs on 5.x: what to do about character devices

[Robert Watson]

On Wed, 4 Jun 2003, Robert Watson wrote:

> It was recently brought to my attention that some users of FreeBSD have
> somewhat populated /compat/linux/dev trees with hard-coded character
> device entries in there.  Often, this seems to occur on upgrded 4.x
> machines running on 5.x The problem with this is that we're in the
> throes of gradually moving away from hard-coded major device numbers, so
> those entries will grow increasingly inaccurate over time.  I'd like to
> find out a bit more about how /dev is managed in linux compat land, what
> ports do special things with it, and how we can make sure that the linux
> compat code runs without any hitches on 5.2-RELEASE when we go -STABLE. 
> 
> >From a 5.x technical perspective, any ideal solution would avoid any
> character device nodes anywhere outside of /dev, so that would mean
> avoiding any entries in a linux-compat-specific dev, or making sure they
> are symlinks to the real /dev, or the like.  Feedback very welcome! 

Just to briefly follow up (sorry, should have been in the earlier e-mail),
linux_base installs a null node, which can probably be ommitted if the
lookup properly falls back to the real system /dev.

Last I checked, vmware depended on rtc, which generated a pseudo-device
(and possibly a node for it?), and on the vmnet bits, which are basically
if_tap, and hence possibly another node.

I'm guessing all of these could simply be "if not -current", or
alternatively "make symlinks".y

Robert N M Watson             FreeBSD Core Team, TrustedBSD Projects
robert at fledge.watson.org      Network Associates Laboratories