Bug 1955 - Using X11 forwarding with SSH for glx commands causes crash
Summary: Using X11 forwarding with SSH for glx commands causes crash
Status: RESOLVED WORKSFORME
Alias: None
Product: xorg
Classification: Unclassified
Component: Server/General (show other bugs)
Version: 6.8.0
Hardware: x86 (IA32) Linux (All)
: high critical
Assignee: Xorg Project Team
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2004-11-28 11:57 UTC by Sterling Stein
Modified: 2007-11-25 12:02 UTC (History)
2 users (show)

See Also:
i915 platform:
i915 features:


Attachments
Simple GL test program (6.48 KB, text/x-csrc)
2007-11-25 12:02 UTC, Antti P Miettinen
no flags Details

Description Sterling Stein 2004-11-28 11:57:21 UTC
I wanted to try sending OpenGL commands over SSH,
so I turned on X11 forwarding in SSH and did
>ssh -Y mymachine
>glxgears

It ran fine for 10 seconds (and printed 2 fps counts)
and then my entire machine locked up.
CTRL+ALT F2 and the power button did not work.

>X -version
X Window System Version 6.8.0
Release Date: 8 September 2004
X Protocol Version 11, Revision 0, Release 6.8
Build Operating System: Linux 2.6.7-gentoo-r6 i686 [ELF]
Current Operating System: Linux s3 2.6.7-gentoo-r6 #1 Wed Jun 30 19:36:27 CDT
2004 i686
Build Date: 20 November 2004
        Before reporting problems, check http://wiki.X.Org
        to make sure that you have the latest version.
Module Loader present

xorg.conf section:
Section "Device"
    Identifier  "Radeon"
    Driver      "radeon"
    #VideoRam    65536
    # Insert Clocks lines here if appropriate

#    Option      "AGPMode" "1"
    Option      "AGPFastWrite"   "yes"
    Option      "NoAccel"        "Off"
    Option      "EnablePageFlip" "On"
EndSection
Section "DRI"
   Mode 0666
EndSection
Comment 1 Roland Mainz 2004-11-30 12:29:36 UTC
Reporter:
Does it help when you turn DRI off ?
Comment 2 Sterling Stein 2004-11-30 12:39:09 UTC
> Does it help when you turn DRI off ?
I did it with DRI on since I wanted to have 3D graphics.
I have not tested it with DRI off and am wary of doing
so since I don't want my computer to crash again.
Comment 3 Adam Jackson 2004-12-01 19:43:41 UTC
you'll still have 3d graphics with DRI off, they'll just be slower.
Comment 4 Roland Mainz 2004-12-02 13:27:19 UTC
ajax@nwnk.net wrote:
> you'll still have 3d graphics with DRI off, they'll just be slower.

It's still accerlated 3D, right ?
Comment 5 Adam Jackson 2004-12-02 13:54:21 UTC
(In reply to comment #4)
> > you'll still have 3d graphics with DRI off, they'll just be slower.
> 
> It's still accerlated 3D, right ?

no.  GL will be slower with DRI off because you'll be rendering everything in
software.  thus, not hardware accelerated at all.
Comment 6 Roland Mainz 2004-12-02 14:08:40 UTC
ajax@nwnk.net wrote:
> (In reply to comment #4)
> > > you'll still have 3d graphics with DRI off, they'll just be slower.
> > 
> > It's still accerlated 3D, right ?
>
> no.  GL will be slower with DRI off because you'll be rendering everything in
> software.  thus, not hardware accelerated at all.

Is there a bugzilla bug filed for that problem yet (Solaris has hardware
accerlation for both "direct" and "indirect" rendering so it doesn't seem to be
a technical limitation :) ?
Comment 7 Adam Jackson 2004-12-02 14:19:56 UTC
there's not a bug about it, but i am quite aware of it and have been working
towards fixing it.

direct versus indirect, in GL, only means whether the client or the server does
the rendering.  it's just that our server doesn't know how to render GL in
hardware yet.
Comment 8 Roland Mainz 2004-12-02 16:01:23 UTC
ajax@nwnk.net wrote:
> there's not a bug about it, but i am quite aware of it and have been working
> towards fixing it.

Is there any ETA when that will be working ?

> direct versus indirect, in GL, only means whether the client or the server 
> does the rendering.  it's just that our server doesn't know how to render GL 
> in hardware yet.

"indirect" rendering may be a advantage in the future when multicore CPUs become
more widespread (Xserver and client can then operate on seperate cores then).
Comment 9 Adam Jackson 2004-12-02 16:15:21 UTC
> Is there any ETA when that will be working ?

no.  to be honest it's been on DRI's todo list since DRI was first written, it's
just that no one has gotten around to it yet.  i'm making headway but i don't
know when i'll have it working.
Comment 10 Adam Jackson 2005-02-01 19:44:00 UTC
unable to reproduce on any of my cards, and a glxgears lockup fix for radeon has
recently been committed.  please reopen if this is still an issue.
Comment 11 Antti P Miettinen 2007-11-25 12:02:09 UTC
Created attachment 12718 [details]
Simple GL test program

I just noticed that I can repeatably (one or two runs of the test program) lock up my X60 tablet running Ubuntu/gutsy by running the attached test program remotely (either tunneled over ssh or not). I tried the same program on another machine (ATI graphics card) and that system is not affected so this might be specific to the "intel" driver. Should there be a separate bug? The hang does not kill the whole machine, the Xorg process is spinning but mouse pointer reacts to movement and I can log in via ssh and kill Xorg, start another etc.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.