Why does loading radeonkms via /boot/loader.conf fail ?

Manish Jain bourne.identity at hotmail.com
Sat Jul 22 17:03:16 UTC 2017

On 07/22/17 20:29, Patrick McMunn wrote:
> I'm pretty sure you can't use both modules; they're mutually exclusive. 
> If you want kms, don't load the radeon module; only load the radeonkms 
> module. Loading radeon first is probably preventing radeonkms from loading.

Hi Patrick,

Thanks a lot for replying. But what you claim - radeon and radeokms are 
likely mutually exclusive - certainly is not correct.

I have booted my box with /etc/rc.local holding :

kldload radeon
kldload radeonkms

That works like a charm. Here is the output of kldstat | egrep 'radeon'

18    1 0xffffffff81b4d000 55b2a    radeon.ko
20    1 0xffffffff81bbb000 11777e   radeonkms.ko
25    1 0xffffffff81d2d000 18f3     radeonkmsfw_CAICOS_pfp.ko
26    1 0xffffffff81d2f000 1cef     radeonkmsfw_CAICOS_me.ko
27    1 0xffffffff81d31000 136e     radeonkmsfw_BTC_rlc.ko
28    1 0xffffffff81d33000 6593     radeonkmsfw_CAICOS_mc.ko

Actually working with vanilla radeon, with no radeonkms around, works 
largely - except that X is limited to low resolution.

There is another problem I face - and I am fairly certain there are 
other members who have experienced this under 10.3 (which defaults to sc 
as the kern.vty driver). startx fails and even 'Xorg -configure' crashes 
with a core - the Xorg log file complains that the node /dev/dri/card0 
is not available. The only way to fix this (AFA I am aware) is to put 
"kern.vty=vt" in /boot/loader.conf and reboot.

There could be a relation between the 2 syndromes - kldloading radeonkms 
should be the right way to create /dev/dri/card0.

So the question expands to :

1) Why can radeonkms not be loaded via /boot/loader.conf ?
2) What makes X dependent on vt as the kern.vty driver ?

Thanks for any pointers.
Manish Jain

More information about the freebsd-questions mailing list