Bug 181736 - connecting to vnc module with X radeon server freezes box
connecting to vnc module with X radeon server freezes box
Product: Fedora
Classification: Fedora
Component: xorg-x11-drv-ati (Show other bugs)
x86_64 Linux
medium Severity medium
: ---
: ---
Assigned To: X/OpenGL Maintenance List
Depends On:
  Show dependency treegraph
Reported: 2006-02-15 23:23 EST by Alexandre Oliva
Modified: 2007-11-30 17:11 EST (History)
1 user (show)

See Also:
Fixed In Version: vnc-4.1.2-18.fc7 kernel- xorg-x11-server-Xorg-
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2007-08-01 14:36:23 EDT
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)
panic when the box freezes (1.31 KB, text/plain)
2006-08-05 12:26 EDT, Alexandre Oliva
no flags Details

  None (edit)
Description Alexandre Oliva 2006-02-15 23:23:05 EST
Description of problem:
See below

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

How reproducible:
Every time

Steps to Reproduce:
1.Set up the X server for a box with an ATI 9200SE card (radeon module) as a vnc
server (Load "vnc")
2.Start the X server
3.Connect to it with vncviewer

Actual results:
After vncviewer opens the window in which the remote screen would be displayed,
it starts painting the window with black, but the box freezes hard part-way
through the process.  Not even SysRq works.

Expected results:
It works fine on other boxes with nv as the primary driver, as well as on this
very box using the vesa driver.

Additional info:
I'm yet to try whether disabling DRI would be of any help.
Comment 1 Alexandre Oliva 2006-02-16 05:26:51 EST
Same error using the "ati" driver, that today's test release configured for me.
Comment 2 Alexandre Oliva 2006-05-27 05:17:52 EDT
This still applies to current rawhide.

I found out that, if I switch to VT1 before connecting, then the box won't
freeze, but VNC won't work either: the VNC client will repaint only part of the
screen before X freezes, and then, most of the times, even after restarting X,
some consistent screen corruption ensues.

X doesn't actually hang in this case, it just keeps repeating the following
lines over and over:

(EE) RADEON(0): RADEONWaitForIdleCP: CP idle -22
(EE) RADEON(0): Idle timed out, resetting engine...
(**) RADEON(0): EngineRestore (32/32)
(EE) RADEON(0): RADEONWaitForIdleCP: CP reset -22
(EE) RADEON(0): RADEONWaitForIdleCP: CP start -22
Comment 3 Alexandre Oliva 2006-06-17 14:05:24 EDT
This is still present in rawhide :-(  It appears to only occur on SMP systems;
with maxcpus=1, the problem did not occur.  Moving the `Load "vnc"' before other
modules enabled vnc to not freeze immediately, but it still froze in similar
ways to those described in 194043, just faster.  Commenting out Load "dri" and
Load "glx" didn't make any visible difference.
Comment 4 Alexandre Oliva 2006-07-27 00:24:04 EDT
With the following combination (rawhide-20060726):


vnc just worked.  Yay!, I thought.  No such luck with the non-xen kernel,
though.  That was because DRI/drm could not be brought up properly in the xen
kernel.  I tried both XAA and EXA AccelMethods, to no avail.  Option "NoAccel"
"on" worked around it, but no other option documented in the man page did, not
even Option "RenderAccel" "off", which I was hoping might fix it.  Another
alternative that now works is to comment out Load "dri".

This is most definitely not a vnc problem.  It might be a problem in the
radeon/dri drivers in X, or in the kernel drm side.  Hard to tell.
Comment 5 Alexandre Oliva 2006-07-28 16:48:30 EDT
Removing the Load "dri" was not enough to get around the entire problem,
unfortunately.  The box still froze after some light use of vnc :-(
Comment 6 Alexandre Oliva 2006-08-05 12:26:16 EDT
Created attachment 133692 [details]
panic when the box freezes

Armed with a serial console, I managed to capture what the kernel printed
before freezing.  It talks about an MCE log, but unfortunately there's no time
to run mcelog before the logged data is history :-(
Comment 7 Matěj Cepl 2007-06-25 09:23:20 EDT
Since this bugzilla report was filed, there have been several major updates,
which may have resolved this issue. Users who have experienced this problem are
encouraged to upgrade their system to the latest version of their distribution

Please, if you experience this problem on the up-to-date system, let us now in
the comment for this bug, or whether the upgraded system works for you.
Comment 8 Matěj Cepl 2007-08-01 14:11:37 EDT
Reporter, could you please confirm as that the bug is still reproducable? If you
won't reply in one month, I will have to close this bug as INSUFFICIENT_DATA.
Thank you.
Comment 9 Alexandre Oliva 2007-08-01 14:36:23 EDT
It *seems* like this is fixed in F7, but I haven't used vnc lately as often or
as heavily as I used to, so I can't quite tell.  It surely doesn't freeze right
away any more.  Thanks,

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