Empty /proc/dri/0/bufs queues -

Michel Dänzer michel at daenzer.net
Sat Aug 13 09:34:54 PDT 2011

On Don, 2011-08-11 at 17:08 -0400, Vipin wrote:
> *I think the first message got discarded, posting a second one*

It didn't, maybe it was delayed via the moderation queue.

> Empty /proc/dri/0 bufs, queues files
> Is this an expected behavior ? 

Yes, these aren't relevant with KMS.

> /sys/kernel/debug/dri/
> There are two directories 0 & 64, what does the 64 entry signify ?

It corresponds to the control device node /dev/dri/controlD64 . There
are legacy/control/render nodes, but I don't remember the details about
the distinction. 

> The contents of /sys/kernel/debug/dri/0 contains 15 radeon_ib_00xx 
> Does this mean, the card has 15 indirect buffers ?

The card imposes no restrictions on the number of indirect buffers (IBs)
that can be allocated. The radeon driver pre-allocates 16 IBs for
dispatching commands to the GPU. 

> Also, does this interface shows me the ringbuffer data like i915.


> The kernel log shows entries like these (massive amounts)
> Aug 11 13:19:01 gilubuntu1 kernel: [ 4251.705377] [drm:drm_ioctl],
> pid=1544, cmd=0xc0086464, nr=0x64, dev 0xe200, auth=1
> Aug 11 13:19:01 gilubuntu1 kernel: [ 4251.705394] [drm:r600_irq_set],
> r600_irq_set: sw int
> Shouldn't it have other debug entries other than these two ?

It should, they might be drowned by these two. Or maybe you need to
enable more debugging messages e.g. with drm.debug=0x3 or =0x7. 

Earthling Michel Dänzer           |                   http://www.amd.com
Libre software enthusiast         |          Debian, X and DRI developer

More information about the dri-devel mailing list