X.org Lockups (part III)

Message

Post by **NeddySeagoon** » Sun May 13, 2007 10:34 pm

dlublink · Post by **dlublink** » Tue May 15, 2007 2:00 am

I had my Xorg using 100% CPU (guess my machine wasn't fast enough to lock up). As per advice from the post of some other users I modified /etc/X11/xorg.conf and disabled the line RenderAccel. I believe this may of resolved the issue for me since Xorg is now using a reasonable percentage of CPU (less than 3%).

Hope this helps someone else :D

David

LXj · Post by **LXj** » Sun May 20, 2007 8:11 am

I upgraded several things, most notably Xorg from 7.1 to 7.2 and started experiencing strange problems with some GTK apps (firefox and freeciv): when I run them, X starts to consume 85-100% cpu

Code: Select all

Section "Module"
    Load        "dbe"   # Double buffer extension
    SubSection  "extmod"
      Option    "omit xfree86-dga"   # don't initialise the DGA extension
    EndSubSection
    Load        "type1"
    Load        "freetype"
    Load       "glx"
    Load       "dri"
EndSection
Section "Files"
    FontPath   "/usr/share/fonts/misc/"
    FontPath   "/usr/share/fonts/TTF/"
    FontPath   "/usr/share/fonts/Type1/"
    FontPath   "/usr/share/fonts/75dpi/"
    FontPath   "/usr/share/fonts/100dpi/"
    FontPath   "/usr/share/fonts/local/"
EndSection
Section "ServerFlags"
EndSection
Section "InputDevice"
    Identifier  "Keyboard1"
    Driver      "kbd"
    Option "AutoRepeat" "500 30"
    Option "XkbModel"  "itouch"
    Option "XkbLayout"  "us,ru(winkeys)"
    Option "XkbOptions" "grp:caps_toggle"
EndSection
Section "InputDevice"
    Identifier  "Mouse1"
    Driver      "mouse"
    Option "Protocol"    "IMPS/2"
    Option "Device"      "/dev/input/mice"
    Option    "ZAxisMapping"    "4 5"
EndSection
Section "Monitor"
    Identifier  "LG Flatron F700P"
    HorizSync   30 - 98
    VertRefresh 50 - 160
EndSection
Section "Device"
    Identifier  "Standard VGA"
    VendorName  "Unknown"
    BoardName   "Unknown"
    Driver     "vga"
EndSection
Section "Device"
    Identifier "ATI Radeon9200"
    BusID       "PCI:1:0:0"
    Option      "AGPMode" "4"
    Driver      "radeon"
    Option      "VideoOverlay" "on"
    Option      "UseFastTLS" "1"
    Option      "EnablePrivateBackZ" "on"
    Option      "RenderAccel" "on"
    Option      "AccelMethod"   "EXA"
EndSection
Section "Extensions"
 Option "XVideo" "Enable"
EndSection
Section "dri"
    Mode        0666
EndSection
Section "Screen"
    Identifier  "Screen 1"
    Device      "ATI Radeon9200"
    Monitor     "LG Flatron F700P"
    DefaultDepth 24
    Subsection "Display"
        Modes       "1024x768" "800x600" "640x480" # "1280x1024" "1152x864"
    EndSubsection
EndSection
Section "ServerLayout"
    Identifier  "Simple Layout"
    Screen "Screen 1"
    InputDevice "Mouse1" "CorePointer"
    InputDevice "Keyboard1" "CoreKeyboard"
EndSection
Section "Extensions"
    Option "RENDER" "Enable"
EndSection

arnuld · Post by **arnuld** » Sun Jun 24, 2007 5:50 pm

but for me these are 2 kinds of problems:

# 1) when running "X" EVERYTHING just freezes. C-M-F2(3,4 rtc) does not work, it is locked, frozen, dead. you can not do anything to solve it the only option is to use "restart" button on Cabinet.

#2) exactly same as #1 but this time mouse moves, i mean mouse pointer does not freeze. but again, the solution is same as the #1.

on Arch, this problems occur 3 times day and on Gentoo it occurs 2-3 times a week. aaarghhh... my friends say that "Linux is same as Windows, it hangs and then you need reboot"

i also had this problem on Arch: http://bbs.archlinux.org/viewtopic.php? ... 87#p261187 but i never had it in Debian, nor i have it in CRUX,the one i use now.

wision · Post by **wision** » Wed Jul 11, 2007 9:39 pm

Hi!
I've got same problem as couple of guys here but noone seems to find a solution.. My xorg freezes when i change CPU frequency while i'm in X, but if i run xorg, switch back to console, change the frequency and go back to X, everything is fine and pc is working on other frequency. I use 2.6.20-r8 with ACPI support, xorg 7.2 and pentium4m as CPU. Any suggestions for solution?

billydv · Post by **billydv** » Sat Jul 14, 2007 3:47 pm

I know this has been a problem for alot of us and its absolutely maddening but I may have found a solution (as crazy as that seems). I never had the problem on my first box, a socket a amd, but when I installed on a socket 939 with an x2 4200 the problems began, I have been experiencing lockups from the time I emerged gentoo-sources and compiled the kernel ,nvidia-drivers. Whats changed from the live-cd kernel to my config, The scheduler, from anticipatory which is the genkernel default I believe to the cfq which I also believe most of us select as desktop users. Now in the last couple of days I have compiled 3 kernels, One with the previous settings, server , cfq sheduler, no prempt kernel,discont memory, timer 250 -- This one freezes
next one has server , anticipatory scheduler, no preempt kernel, discont memory, timer 250 --no freezes left overnight nor any delays when resuming from screensaver

next one has desktop (not the low latency), anticipatory scheduler, preempt kernel, sparse memory, and timer 1000 --no freezes for 6hrs and I just did a test, copied 9 individual iso,s simultaneously over samba shares, ran revdep-rebuild and emerge glibc and all copies completed and revdep-rebuild completed with no problems before I canceled the glibc emerge and no problems, no need to restartx

I admit I know very little about all of this compared to many of the people having problems but look at the people on the nvidia forums that are having problems, the majority use distros where you compile your kernel (in ex gentoo) or ubuntu (default cfq scheduler). Its not long enough to be conclusive but I know that my results are so far repeatable. Please inform of any results you may come up with.

Joaquin24 · Post by **Joaquin24** » Wed Jul 25, 2007 4:23 am

Hi, just to prove my testimonial

compiled kernel like 30 times, varying different versions from 2.16.6 to 2.6.20 with desktop patchset's,

I have two machines one AM2 +3000, but that one didn't gave me so much trouble, however I didn't not run it heavily, as this has an early nforce mobo, the only kernel that supports this are 2.6.21 or higher so no quirks nor complain on that one, on the secondary Pc is where I ve done most compilations.

next one has server , anticipatory scheduler, no preempt kernel, discont memory, timer 250 --no freezes left overnight nor any delays when resuming from screensaver

this type of server approach gave me the best results, nice to proof concept over here

.

and also

Option "RenderAccel" "false"

gave me best results.

Both pc's are running nvidia GPU's. I've gone through some kernel parameters to spare memory to for applications (X, Desktop, etc)

Code: Select all

vm.overcommit_ratio = 50
vm.overcommit_memory = 1
vm.drop_caches = 3

simple one, but effective, free ups 20~30 from the disk cache, thus physical ram for apps to play around.

Now

at least I know that having a server kernel approach, over a desktop it's not absolutly nuts..

billydv · Post by **billydv** » Wed Jul 25, 2007 10:14 am

Joaqin, I am curious as to what the kernel parameters that you set are, the ones that you mention in the previous post, the vm stuff for the nvidia cards? They arent in the kernel config, at least not that I can see.

billydv · Post by **billydv** » Wed Jul 25, 2007 10:38 am

Okay, I see that you can make the changes via the sysctl file in etc but what effect do you think these changes will have towards making xorg freeze less?

Joaquin24 · Post by **Joaquin24** » Sun Jul 29, 2007 2:18 am

the vm stuff for the nvidia cards?

no there is not vm stuff for nvidia, that's sysctl.conf settings

Code: Select all

man sysctl.conf

I 'm not an expert, but it's my favourite tool, you can search over the kernel mailing list to read over what does what, you can do pretty much everything when speaking of memory with sysctl.

speaking about overcommit http://www.uwsg.iu.edu/hypermail/linux/ ... /0117.html
http://lkml.org/lkml/2007/4/10/180

but if you have absolutly no clue of what you are doing over sysctl, you can scr3w up, (kernel pacnic )

Okay, I see that you can make the changes via the sysctl file in etc but what effect do you think these changes will have towards making xorg freeze less?

well if I understood correct is a setting that tweaks the cache of the memory, and allows to resctrict the writing over and over of certain applications over the disk cache memory, I don't know if it helps out with Xorg, maybe it will not help out with indexing files, it some kind of conservative setting , don't know it's long to explain, it just does what I want.

(edit) correction not for embebbed systems, here is nice explanation by Alan cox

It deals with
the theoretical worst cases and ensure that a page write won't cause an
out of memory situation. That means it is more pessimistic than the usual
workload. It also by default allows lots of room for worst case kernel
behaviour that is appropriate to a real system with swap but not embedded

update
so my reasoning is that, pushin always the memory to free up, does improve responsiveness on Xorg applications, having more raw memory helps out(this is not the same as swappiness)
though that will not make Xorg undefeat-able , there is always a chance that a cr4p programmed aplication lockups the entire X, or the whole system in some cases,
And worst case, with this settings the magical SysRq showed more responsive.

well, I m running with 256 dimm PC133, Nvidia Geforce 2, (have to take care of Ram)
it proved to be much more responsive

(sorry if the post doesn't make sense as reading as a whole, i made corrections several times)

cheers

Joako.-

> I m interested in more Xorg cheap hacks... does anyone knows any other? <

billydv · Post by **billydv** » Mon Jul 30, 2007 10:58 pm

Okay, as I have been having the xorg lockups for a while and its a variable thing, sometimes it does and sometimes it doesnt, for starters, I realize I had alot of other small problems going on using up memory
1- A bad network cable was causing dhcpcd and syslog-ng to run nonstop and heavy cpu usage

2- Xorg.conf had numerous mistakes, I didnt realize until a kernel recompile was causing the keyboard to act funny, I ran X- configure and used that as a base to redo my xorg.conf

We'll see if it made much of a difference.

Joaquin24 · Post by **Joaquin24** » Sun Aug 05, 2007 2:07 am

I realize I had alot of other small problems going on using up memory

yeah, I dont know what's the buggy sh7t, but I can tell the thing is not getting better with much memory, running with 1GB ram, and an Am2, was same, I wanted to get a re-fund (because it was brand new), but manage to to underclock some settings on the bios, and didnt wanted to go on burocratics
I m 98 percent sure that something buggy on the nvidia drivers, come on!!

, is there any precedent on this, I turned the box on and the glitches where so big that I almost got blind.

At least we know where the problem is at the nvidia hardware GPU, they get super hot, and then the glitches, artifacts etc lockups, start showing up.

Some time ago I plugged a fan from a HP laserjet 6

, it turned out to be the best solution whatsover

All this nvidia's were ment for DirectX way crappy games, that turn on and off, not for continous heavy load..

Now if somehow nvidia releases a power manager or something that limits the GPU stressing,

cillianderoiste · Post by **cillianderoiste** » Tue Aug 07, 2007 8:02 pm

I have experienced a lot of lock ups too, but have managed to fix them eventually. I've had pretty much the same issues before with another nvidia card and I have one system which never has these issues. To check if it's an issue with the nvidia driver you can look at your var/log/messages for Xid errors:
grep Xid /var/log/messages
Aug 6 15:17:06 localhost NVRM: Xid (0002:00): 16, Head 00000000 Count 00058a68
Aug 6 15:17:07 localhost NVRM: Xid (0002:00): 8, Channel 00000000
Aug 6 15:17:07 localhost NVRM: Xid (0002:00): 30, L1 -> L0
Aug 6 15:17:23 localhost NVRM: Xid (0002:00): 8, Channel 0000001e
Aug 6 15:17:23 localhost NVRM: Xid (0002:00): 30, L0 -> L0
Aug 6 15:17:24 localhost NVRM: Xid (0002:00): 16, Head 00000000 Count 00058c49
Aug 6 15:17:31 localhost NVRM: Xid (0002:00): 8, Channel 00000020
Aug 6 15:17:32 localhost NVRM: Xid (0002:00): 16, Head 00000000 Count 00058c4a
Aug 6 15:17:39 localhost NVRM: Xid (0002:00): 8, Channel 00000020
I have tried many many things to resolve it and I think it's stable now. The best thing is to check the nvidia forums for advice. What made the difference for me was having idle=poll as a kernel parameter (recommended for dual core, mine is amd64 X2). I rather forcefully disabled agpgart from the kernel but that didn't seem to make any difference.
Since my system is a laptop I also have:
in
/etc/modules.d/nvidia
options nvidia NVreg_RegistryDwords="PerfLevelSrc=0x2222" NVreg_Mobile=3

I think that disables a powersaving mode

and in xorg.conf:
Option "NvAGP" "1"

which is supposed to force nvidia's agp

These steps are not likely to work with every system so check the forums. I also had severe problems with 100.14.11 but 100.14.09 is fine and I run compiz-fusion as my window manager without further problems. Incidentally the problems I had most recently also occurred even with all hardware acceleration turned off and with a plain metacity session.

If you can log into your machine via ssh after it locks up and run nvidia-bug-report.sh that will give you lots of details. The best way to get help is to post this along with details of the error to the forums on http://www.nvnews.net. You may get a reply from someone in nvidia but you'll probably start off with lots of awkward options to try out. Nvidia's initial response will be to update your bios and kernel, although I don't believe I've come across anyone who that has worked for.

It is always an extremely annoying problem when you come up against it, on one system I gave up and bought another card which worked fine. Of course that isn't an option on a laptop.

Good luck!

DancesWithWords · Post by **DancesWithWords** » Wed Aug 08, 2007 1:37 am

NeddySeagoon wrote:Contines This thread

For the last while I've been experiencing complete lock-ups of by computer.

I've running
Dual AMD 246 Opterons
1Gb Ram
Sparkle Nvidia 7600, 512MB
Nvidia Module 100.14.09
Kernel 2.6.21-r4

I've tried everything. I can think of to fix it. The only I've seem that helps is using the NV module. This is quite madding. So it must be related to the Nvidia module. Has anyone got a fix. I've just tried Kernels 2.6.17.-r4, 2.6.20-r8 and I still get the same problem.

--
DWW

DancesWithWords · Post by **DancesWithWords** » Wed Aug 08, 2007 1:39 am

NeddySeagoon wrote:Contines This thread

For the last while I've been experiencing complete lock-ups of by computer.

I've running
Dual AMD 246 Opterons
1Gb Ram
Sparkle Nvidia 7600, 512MB
Nvidia Module 100.14.09
Kernel 2.6.21-r4

I've tried everything. I can think of to fix it. The only I've seem that helps is using the NV module. This is quite madding. So it must be related to the Nvidia module. Has anyone got a fix. I've just tried Kernels 2.6.17.-r4, 2.6.20-r8 and I still get the same problem.

Oh and I notice that most of my crashes occur when playing games, kstars, and google earth.

--
DWW

Nerdanel · Post by **Nerdanel** » Thu Aug 09, 2007 5:53 pm

As a person who has suffered considerable trouble with these lockups and tried pretty much everything, before finally discovering the real cause to be something I never thought of.

It may be your power source failing gradually.

For me the freezes appeared after I had replaced my motherboard. I suspected everything from coincident software updates/recompiles to BIOS settings to the northbridge fan providing inadequate cooling. I got rid of my problems for several months by replacing the northbridge fan with a better-conducting (and passive) one, but now I see it worked only because it reduced my computer's power use by one active fan. And after a few months the lockups were back worse than ever.

My Antec power source had probably been failing since the beginning, but it had taken time and a chance to a new motherboard until the power level had decreased enough to cause problems. The final clue was that I started to experience problems that my motherboard manual were usually caused by overclocking too much... but I wasn't overclocking at all. My system has been 100% stable for over six months now with a new power source, so I can definitely say the power source is something you need to keep in mind.

DancesWithWords · Post by **DancesWithWords** » Fri Aug 10, 2007 12:23 am

Nerdanel wrote:As a person who has suffered considerable trouble with these lockups and tried pretty much everything, before finally discovering the real cause to be something I never thought of.

It may be your power source failing gradually.

For me the freezes appeared after I had replaced my motherboard. I suspected everything from coincident software updates/recompiles to BIOS settings to the northbridge fan providing inadequate cooling. I got rid of my problems for several months by replacing the northbridge fan with a better-conducting (and passive) one, but now I see it worked only because it reduced my computer's power use by one active fan. And after a few months the lockups were back worse than ever.

My Antec power source had probably been failing since the beginning, but it had taken time and a chance to a new motherboard until the power level had decreased enough to cause problems. The final clue was that I started to experience problems that my motherboard manual were usually caused by overclocking too much... but I wasn't overclocking at all. My system has been 100% stable for over six months now with a new power source, so I can definitely say the power source is something you need to keep in mind.

I had similar problems with a Enermax 550 watt power supply. So I upgraded to a 650 watt quad rail power supply. Bu this power supply is maybe a year old I have a hard time thinking this could be the source of the problem. I hope that is not the issue I'm not willing at this stage to drop another $285.00 on another PS.

I'll try downgrading Xorg to 7.1 and the Nvidia drivers to see if that helps.

--
DWW

DarkMetatron · Post by **DarkMetatron** » Fri Aug 10, 2007 7:24 pm

I only have this lockups when I use KDE. Now I use XFCE4 and the lockups are gone for good.

transient · Post by **transient** » Sat Aug 11, 2007 10:35 am

DancesWithWords wrote:
NeddySeagoon wrote:Contines This thread
For the last while I've been experiencing complete lock-ups of by computer.

I've running
Dual AMD 246 Opterons
1Gb Ram
Sparkle Nvidia 7600, 512MB
Nvidia Module 100.14.09
Kernel 2.6.21-r4

I've tried everything. I can think of to fix it. The only I've seem that helps is using the NV module. This is quite madding. So it must be related to the Nvidia module. Has anyone got a fix. I've just tried Kernels 2.6.17.-r4, 2.6.20-r8 and I still get the same problem.

Oh and I notice that most of my crashes occur when playing games, kstars, and google earth.

--
DWW

The fact they mostly occur when you're under heavy video card usage lends more weight to it indeed being your PSU. Not necessarily that it's failing, but it might just not be able to provide enough power to all your components at the same time.

DancesWithWords · Post by **DancesWithWords** » Sun Aug 12, 2007 4:46 am

transient wrote:
DancesWithWords wrote:
NeddySeagoon wrote:Contines This thread
For the last while I've been experiencing complete lock-ups of by computer.

I've running
Dual AMD 246 Opterons
1Gb Ram
Sparkle Nvidia 7600, 512MB
Nvidia Module 100.14.09
Kernel 2.6.21-r4

I've tried everything. I can think of to fix it. The only I've seem that helps is using the NV module. This is quite madding. So it must be related to the Nvidia module. Has anyone got a fix. I've just tried Kernels 2.6.17.-r4, 2.6.20-r8 and I still get the same problem.

Oh and I notice that most of my crashes occur when playing games, kstars, and google earth.

--
DWW
The fact they mostly occur when you're under heavy video card usage lends more weight to it indeed being your PSU. Not necessarily that it's failing, but it might just not be able to provide enough power to all your components at the same time.

Okay I'll test that out. I'll get another PSU and and use it only to power the video card and see what that does. Funny I was coming to a similar conclusion. I won't have one until Wednesday I'll post my results.

I've made a few other alterations to up set since the last time to try and fix things. In the kernel -> I've got a Preemption Model (Preemptible Kernel (Low-Latency Desktop)) and I changed Timer frequency (1000 HZ). Lastly I've downgrade my Nvidia drivers to 1.0.8776-r1

--
DWW

Ainulindalë · Post by **Ainulindalë** » Mon Aug 13, 2007 6:06 am

I'm experiencing pretty much the same thing, with this configuration :

AMD64 3000
Ati Radeon 9200
Kernel 2.6.22-r2

I'm in ~amd64, and it worked pretty nicely until I updated the kernel to the .22 release. I think it's possible that at the time of the update, I switched from anticipatory to CFQ scheduler (which is the present state). I will try to change the option and see how it goes. I remember pretty surely having some problems with CFQ on earlier versions of the kernel, problems which forced me to switch scheduler.

Right now, whereas my configuration hasn't change a bit, it locks up on trivial operations (using a drop down menu), without any apparent reason, at unforeseeable times. It simply goes to 90-100% CPU usage, sometimes the mouse pointer is able to move, sometimes not, but it's usable by SSH. Though, it's impossible to kill xorg or to restart it, as it's in zombie state.

I'm not keen on the PSU hypothesis, as for my part, my system isn't that much demanding (two hard drives, one DVD drive, and an old video card).

st0ned · Post by **st0ned** » Mon Aug 13, 2007 10:14 pm

My system is now stable after adding maxcpus=1 as a kernel boot parameter. This is less than ideal however as I would prefer to make use of both CPUs

maartenlambrecht · Post by **maartenlambrecht** » Tue Aug 14, 2007 2:33 pm

I think my topic is related to this one. It's located here: http://forums.gentoo.org/viewtopic-t-575841.html
The error I get after running Xorg-x11 with or withouth kde is this one:

Code: Select all

FreeFontPath: FPE "/usr/share/fonts/misc/" refcount is 2, should be 1; fixing.

This is the reason KDE doesn't work, so I would like to know a solotion.

Ainulindalë · Post by **Ainulindalë** » Fri Aug 17, 2007 6:38 am

I tried yesterday to change from CFQ to Anticipatory : same result, it freezes (meaning Xorg goes to 100% CPU, mouse is moving but keyboard isn't usable) without any intervention, or pretty quickly if I try to use graphical widgets.

For example, last time it freezed was when I tried to launch a program with the basic menu of XFCE. It freezed while painting this menu.

I tried to append noapic acpi=off idle=poll to my kernel configuration yesterday night, and switched to 2.6.22-r3. I also pointed out that I had useless options in my .config : multiprocessing, support for multi-core (and multi-core scheduler), though I use a single core processor, which I removed. I'll see if it changes anything, but what I find strange is that this bug only showed itself when I migrated from 2.6.21 to 2.6.22.

VinzC · Post by **VinzC** » Fri Aug 17, 2007 10:38 pm

Hi.

Don't know if it's the right place for posting but here I am. I've run through the big upgrade expat+KDE 3.5.5 -> KDE 3.5.7. Now I'm facing strange lockups of X, i.e. it eats 100% CPU after a certain amount of time running the screensaver. The more I leave my machine inactive (with the screen saver running) the longer X consumes CPU but only *after* I unlock the screen.

I've been able to unlock the screen a couple of times since the last time it ate up CPU like this, maybe four or five times but I always noticed CPU activity was still high for a noticeable period of time - a few seconds, maybe 10-20 seconds. And when it locks up I can still switch to a text console and kill -9 X, which is my only option. Strangely enough I can see activity in Superkaramba applets (Aero-IO panel) and the clock in kicker ticking but I can't interact with the desktop.

The first time I saw X lock up I tried zapping it with Ctrl-Alt-BkSpace but then I couldn't even switch to a text console. I've had to SSH into the machine and kill X manually. Running /etc/init.d/xdm stop doesn't work either in such a case.

Does anyone have an idea on that really annoying problem?

X.org Lockups (part III)

X.org Lockups (part III)

Re: X.org Lockups (part III)

Re: X.org Lockups (part III)

Re: X.org Lockups (part III)

Re: X.org Lockups (part III)

X eats 100% CPU after running screensaver for a long period