New AMD drm code
Adam K Kirchhoff
akirchhoff135014 at comcast.net
Sun Mar 17 22:22:23 UTC 2013
On 03/14/2013 05:59 PM, J.R. Oldroyd wrote:
> On Mon, 11 Mar 2013 16:13:45 -0400 Adam K Kirchhoff <akirchhoff135014 at comcast.net> wrote:
>> Thanks, that got it working. I haven't had a chance to actually test
>> anything that uses DRI, but the module loads:
>>
> Sorry, been a bit busy this week.
>
> There are a bunch of of memory leaks in the code as it is at
> the moment. I suspect you have noticed that the server gets slow
> and eventually stops. If you haven't, try opening two xterms.
> Run "top" in one. Run "ls -l /usr/local/bin" a few times in the
> other. Observe the Wired and Free counts in the top display.
> (It's probably a good idea to reboot before either runs out.)
>
> Earlier this week I fixed all the Wired page leaks and sent
> that patch to dumbbell at . It's not committed yet, though.
>
> I have also just fixed a bunch of places where a null pointer is
> freed and I've fixed a couple of the Free page leaks too. I'll be
> sending an updated patch to dumbbell tomorrow evening.
>
> With any luck, dumbbell will have time to commit some of these
> patches to the git repo soon. Maybe this weekend.
>
> There are fifteen major leaks remaining, although some of these
> may be related to one another. I.e., there may be <= 15 bugs
> still to find.
>
> It'd be great to have others help track some of these leaks down.
> My time will be limited next week. The codepaths here are rather
> convoluted, so it's a bit time-consuming. I have a lot of trace
> info about which mallocs are being leaked that I can send to
> anyone who can help.
>
> -jr
Well I'm happy to say that X works for me. It starts up, DRI is
enabled. Sadly, I only have one functional monitor, according to
xrandr, and even that one doesn't report the full 1600x900 resolution
that it supports. xrandr maxes out at 1024x768.
http://adam-kirchhoff.com/Xorg.0.log
http://adam-kirchhoff.com/dmesg.txt
DVI-0 is connected but, as you can see in the log file, isn't detected.
DVI-1 is at least detected :-)
I saw this morning on https://wiki.freebsd.org/AMD_GPU that "EDID is now
read properly so the correct resolution is used by X.Org." So I made
sure to pull the latest code on the kms-radeon branch from
https://github.com/dumbbell/freebsd today and updated the kernel before
reporting this here.
Adam
More information about the freebsd-x11
mailing list