Bug 124653 - X busy loops with ATI Radeon 9800 XT
X busy loops with ATI Radeon 9800 XT
Product: Fedora
Classification: Fedora
Component: hwdata (Show other bugs)
x86_64 Linux
medium Severity medium
: ---
: ---
Assigned To: Bill Nottingham
David Lawrence
Depends On:
  Show dependency treegraph
Reported: 2004-05-28 04:40 EDT by Tim Waugh
Modified: 2014-03-16 22:45 EDT (History)
2 users (show)

See Also:
Fixed In Version: hwdata-0.123-1
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2004-07-09 12:36:02 EDT
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)

  None (edit)
Description Tim Waugh 2004-05-28 04:40:29 EDT
Description of problem:
After installing FC2 from CD onto this AMD64 Opteron, everything goes
fine until after the end of the firstboot screen, when X tries to
start.  The screen goes blank and nothing else happens.

Logging into a text console and running 'ps' I can see that X is in R
state, and 'strace' and 'ltrace' show no output.

The funny thing is that anaconda and firstboot worked fine in
graphical mode.

Version-Release number of selected component (if applicable):

How reproducible:
Comment 1 Tim Waugh 2004-05-28 04:54:02 EDT
If I edit the xorg.conf file and change 'vesa' to 'radeon' it all
works fine (and fast).  Console switching also works.
Comment 2 Tim Waugh 2004-07-09 07:48:11 EDT
With today's rawhide (first I've tested this in weeks) it is even
worse: X doesn't even start properly in Anaconda.  Just get a black

Comment 3 Mike A. Harris 2004-07-09 09:28:43 EDT
The kernel is most likely breaking vesa BIOS access, as nothing
has changed in X.

Nothing I can do about that, but you could file a kernel bug report
for it.

If "radeon" works as it should though, then this is just a hardware
database issue to resolve by pointing the PCI ID to the "radeon"
driver.  I plan PCI ID updates for sometime in August when I
return from vacation.  If you supply lspci -vvn and reassign to
hwdata, someone might change this one card before then for you
though, or you could update it yourself if there's no ACL.

Comment 4 Tim Waugh 2004-07-09 09:37:50 EDT
"radeon" works fine, just slowly.

02:00.1 Class 0380: 1002:4e6a
        Subsystem: 1462:9561
        Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-
ParErr- Stepping+ SERR- FastB2B-
        Status: Cap+ 66Mhz+ UDF- FastB2B+ ParErr- DEVSEL=medium
>TAbort- <TAbort- <MAbort- >SERR- <PERR-
        Latency: 64 (2000ns min), Cache Line Size 08
        Region 0: Memory at d8000000 (32-bit, prefetchable)
        Region 1: Memory at ff5e0000 (32-bit, non-prefetchable) [size=64K]
        Capabilities: [50] Power Management version 2
                Flags: PMEClk- DSI- D1+ D2+ AuxCurrent=0mA
                Status: D0 PME-Enable- DSel=0 DScale=0 PME-
Comment 5 Mike A. Harris 2004-07-09 09:50:32 EDT
Can you attach the X server log, /var/log/messages and the output
of /proc/mtrr?  You're probably getting broken MTRRs.  There is a
kernel bug with MTRR on x86_64 (glibc-kernheaders actually), but
I don't know if/when it gets fixed.  X probably needs to be
recompiled for the fix to get used.

In the mean time, if MTRR isn't active, try setting it manually
via the /proc/mtrr interface.  Your video RAM is at 0xd8000000
so set up an MTRR there with range the size of video RAM.  If
it fails to set it up, you might need to split MTRR which is a
PITA.  Once set up tho, you can script it in rc.local for now
until the kernel header issue is fixed.

Comment 6 Mike A. Harris 2004-07-09 12:36:02 EDT
hwdata-0.123-1 I've added a bunch of radeon entries and mapped them
to radeon driver.  There are probably a lot of cosmetic glitches
and incorrect card names, etc. as I just went by the sf.net data,
which is often wrong.  This should get all ATI hardware we know
about at least pointed to the "radeon" driver now though, and I 
can worry about completing the list and fixing up cosmetics in
a month or so.

Closing as fixed in rawhide for now, but there's another bug
open for the MTRR issues on x86_64.. don't have bug ID handy, but
CC yourself on it and add the info requested above.


Note You need to log in before you can comment on or make changes to this bug.