major number leak with modules?

[Poul-Henning Kamp]

In message <20031207173959.GE42518 at cicely12.cicely.de>, Bernd Walter writes:
>The situation was the following during driver development.
>It's an USB driver and kldload'ed.
>A plugged in device got major 247 for the nodes it created.
>On unplugging the nodes were destroyed.
>kldunloading the driver and kldloading the next revision created
>nodes with major 246 for new devices.
>
>Do we have a leak with major numbers or is the old major number free
>after last destroy_dev and assigning algorithm just took the next.

Yes, repeatedly loading/unloading will leak majors.

I have some ref-counting code to solve this problem.  Warners
axe-swinging in the old ISA drivers made it easier, but some necessary
but uncommitted patches to the sound code from cg@ are still at
road-block.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.