[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: fglrx driver / 1024mb ram
- From: "Michel Dänzer" <daenzer at debian dot org>
- Date: Mon, 04 Apr 2005 10:56:10 -0400
On Sat, 2005-04-02 at 03:11 +0200, Martin Jordan wrote:
the xfree86 startup with 512mb of ram (works very well):
http://fanti.staff.spin.de/xfree86.log_512mb
and exactly the same configuration (same kernel with 4gb highmem support) but
1024mb of ram instead of 512mb:
http://fanti.staff.spin.de/xfree86.log_1024mb
as you can see, xfree stops here:
(II) fglrx(0): [DRI] installation complete
that seems to be the point where xfree freezes (99% cpu usage, just shows a
black screen - and nothing more).
the next line "(II) fglrx(0): Direct rendering enabled" will never be printed
out to the log with 1024mb of ram. i've no idea what's going one here?!?
That's the point where the driver tries to start the card's bus
mastering engine, apparently something goes wrong with that.
fanti at daniela:~$ diff xfree86.log_512mb xfree86.log_1024mb
[...]
650,651c650,651
< (II) fglrx(0): [drm] added 8192 byte SAREA at 0xe0988000
< (II) fglrx(0): [drm] mapped SAREA 0xe0988000 to 0xb7db1000
---
(II) fglrx(0): [drm] added 8192 byte SAREA at 0xf8988000
(II) fglrx(0): [drm] mapped SAREA 0xf8988000 to 0xb7db1000
672c672
< (II) fglrx(0): [drm] texture shared area handle = 0xe4eaf000
---
(II) fglrx(0): [drm] texture shared area handle = 0xfceaf000
[...]
=========
the only differences except the "hangup point" from both logs are these
addresses. might that be useful for troubleshooting?
No, these aren't expected to be constant.
kernel .config: http://fanti.staff.spin.de/config-2.6.11.5
Anything interesting in the kernel output? FWIW, on the PCIe machine
with 1 GB RAM here, HIGHMEM64G is enabled but HIGHMEM4G disabled, but
that's with an RHEL 2.4 kernel, so I'm not sure it's relevant.
--
Earthling Michel Dänzer | Debian (powerpc), X and DRI developer
Libre software enthusiast | http://svcs.affero.net/rm.php?r=daenzer