radeon/amd64 random hangs

Please attach the following files:

/etc/X11/XF86Config-4
/var/log/XFree86.0.log
/var/log/dmesg

Created an attachment (id=647)
XFree86 Log

Created an attachment (id=648)
DMesg

Created an attachment (id=649)
XF86Config

You seem to be running an old kernel; presumably you installed a pre-release.
You can find instructions for bringing your kernel up-to-date here:

http://www.ubuntulinux.org/wiki/WartyWarthogUpgradeNotes

I don't think that will fix your problem, but it is a good idea regardless.

I'll have to second mdz's guess -- does it work with a newer kernel?

I've done an upgrade, and a dist-upgrade and yet I don't see a new kernel?

The newest one I could manually find through apt-cache search was
2.6.8.1-3-amd64-generic...

Am I missing something?

(In reply to comment #7)
> I've done an upgrade, and a dist-upgrade and yet I don't see a new kernel?
>
> The newest one I could manually find through apt-cache search was
> 2.6.8.1-3-amd64-generic...

Correct.

> Am I missing something?

The dmesg that you attached showed that you are not running that kernel:

Linux version 2.6.8.1-2-amd64-generic (buildd@crested) (gcc version 3.3.4
(Debian 1:3.3.4-9ubuntu5)) #1 Sat Sep 18 08:40:04 UTC 2004

Ok, I'll try that one then...

By the way, why did I have to manually install the new kernel, why wasn't it an
automatic upgrade when using dist-upgrade or upgrade?

(In reply to comment #10)
> By the way, why did I have to manually install the new kernel, why wasn't it an
> automatic upgrade when using dist-upgrade or upgrade?

Because you installed a pre-release version. Instructions were posted to the
mailing list when the final release went out:

http://www.ubuntulinux.org/wiki/WartyWarthogUpgradeNotes

Ah, didn't see that. Thanks so much!

Okay, I performed all the upgrade steps listed in that Wiki document, and I'm
now running the latest available kernel from apt-get.

However, this morning about 5 minutes after I booted my system, I was browing
www.cnn.com/ELECTION/2004 in FireFox when the system hardlocked again.

So, it doesn't look like the kernel update fixed it.

Any other suggestions?

I don't want to point fingers, but the last three times it's happened now have
been while using Firefox (0.9.3 here).

I don't know what else to try other than not using Firefox at this point...

Any other suggestions?

I'm pretty sure these lockups are fixed in X.Org.

Not fixed by X.org...

I hadn't been using Ubuntu for several months now because of this bug and
decided to give it another whirl with the Hoary preview release.

I was running the software update manager when the exact same thing that's
happened in the past happened again. Hard lock, machine won't respond to anything.

This issue is specific to Ubuntu, and does not occur with any of the following
operating systems:
RedHat Linux 7.3 (32-bit)
RedHat Linux 8.0 (32-bit)
RedHat Linux 9 (32-bit)
RedHat Enterprise Linux 3 WS Update 3 (32-bit)
Fedora Core 3 (32-bit)
Fedora Core 3 (64-bit)
SuSE Linux 9.2 (64-bit)
Novell Linux Desktop 9 (32-bit)
DragonFlyBSD (32-bit)
Solaris 10 (64-bit)
Windows XP SP2 (32-bit)

There is something specific to how Ubuntu is either patching the kernel or the X
server that is causing problems on my system. No other distribution or OS has
this issue.

Is there any way I can get some kind of information or logging going that would
help an Ubuntu developer track down why this is happening?

Is there any further information I can provide to help?

I forgot to note, as before this is with a clean install of Ubuntu. No evil
binary NVidia or ATi drivers or any of the like. Just plain Ubuntu on a reiserfs
root partition with an ext3 home partition.

I just tried reinstalling the system, this time using ext3 as the root file
system. Same problem, so it's not a reiserfs thing.

Will try ATi's video drivers next.

Weird. Works OK for me with my amd64 and an X300 SE; let me know if you turn
anything up.

Hang happens with ATi's binary drivers as well.

Either this is a deep Ubuntu Xorg problem, or something wrong in the kernel itself.

Going to try 32-bit Ubuntu next. If it works, then you know this is an AMD64
specific Ubuntu problem.

Okay, it's definitely an amd64 specific Ubuntu issue. No problems so far with
the 32-bit preview version of Ubuntu.

Looks as though I spoke too soon, using the 2.6.10-4 kernel (32-bit Ubuntu) after a few hours it
finally hard locked (when I was opening the KDE help center (having just installed
kubuntu-desktop)). It locked up much quicker consistently before running the 64-bit ubuntu
distribution.

I'm now using the xorg-fglrx driver and the latest Ubuntu 2.6.10-5 32-bit kernel. I'll report
back with my findings of this new combination.

Wow. Updating to 2.6.10-5 kernel actually made things worse. Wonderful.

Only idea I have left is maybe the network driver for my realtek onboard Gigabit
ethernet?

This is pretty frustrating, is there any type of logging program or extra
debugging info I can get that will help track this thing down?

Not that I know of. You're not booting with 'inotify', are you? I've been
using either an X300 SE or an X850 XT PE (rv370 and r480, respectively), in this
amd64 for quite some time now, and not seen any random hangs. I suspect that
something is up with a kernel driver here: attaching the full output of 'lspci'
might be useful. The fact that it's 64-bit-specific makes me very, very
suspicious of any X issues; almost all of them would also manifest in 32-bit mode.

I'm doing the default Ubuntu install, I'm not even certain what inotify is.

As far as being 64-bit specific, I originally thought so too, but if you look at
comment #22, you'll see that I had the problem with the 32-bit version of Ubuntu
too (it just took longer).

I'll post the lspci output later today.

00:00.0 Host bridge: nVidia Corporation nForce3 Host Bridge (rev a4)
        Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr-
Stepping- SERR+ FastB2B-
        Status: Cap+ 66Mhz+ UDF- FastB2B+ ParErr- DEVSEL=fast >TAbort- <TAbort-
<MAbort- >SERR- <PERR-
        Latency: 0
        Region 0: Memory at e0000000 (32-bit, prefetchable) [size=128M]
        Capabilities: [44] HyperTransport: Slave or Primary Interface
                Command: BaseUnitID=0 UnitCnt=12 MastHost- DefDir- DUL-
                Link Control 0: CFlE+ CST- CFE- <LkFail- Init+ EOC- TXO-
<CRCErr=0 IsocEn- LSEn- ExtCTL- 64b-
                Link Config 0: MLWI=16bit DwFcIn- MLWO=8bit DwFcOut- LWI=16bit
DwFcInEn- LWO=8bit DwFcOutEn-
                Link Control 1: CFlE- CST- CFE- <LkFail+ Init- EOC+ TXO+
<CRCErr=0 IsocEn- LSEn- ExtCTL- 64b-
                Link Config 1: MLWI=8bit DwFcIn- MLWO=8bit DwFcOut- LWI=8bit
DwFcInEn- LWO=8bit DwFcOutEn-
                Revision ID: 1.03
                Link Frequency 0: 600MHz
                Link Error 0: <Prot- <Ovfl- <EOC- CTLTm-
                Link Frequency Capability 0: 200MHz+ 300MHz+ 400MHz+ 500MHz+
600MHz+ 800MHz- 1.0GHz- 1.2GHz- 1.4GHz- 1.6GHz- Vend-
                Feature Capability: IsocFC+ LDTSTOP+ CRCTM- ECTLT- 64bA- UIDRD-
                Link Frequency 1: 200MHz
                Link Error 1: <Prot- <Ovfl- <EOC- CTLTm-
                Link Frequency Capability 1: 200MHz- 300MHz- 400MHz- 500MHz-
600MHz- 800MHz- 1.0GHz- 1.2GHz- 1.4GHz- 1.6GHz- Vend-
                Error Handling: PFlE- OFlE- PFE- OFE- EOCFE- RFE- CRCFE- SERRFE-
CF- RE- PNFE- ONFE- EOCNFE- RNFE- CRCNFE- SERRNFE-
                Prefetchable memory behind bridge Upper: 00-00
                Bus Number: 00
        Capabilities: [c0] AGP version 3.0
                Status: RQ=32 Iso- ArqSz=2 Cal=0 SBA+ ITACoh- GART64- HTrans-
64bit- FW+ AGP3+ Rate=x4,x8
                Command: RQ=1 ArqSz=0 Cal=0 SBA+ AGP+ GART64- 64bit- FW+ Rate=x8

00:01.0 ISA bridge: nVidia Corporation nForce3 LPC Bridge (rev a6)
        Subsystem: Giga-byte Technology: Unknown device 0c11
        Control: I/O+ Mem+ BusMaster+ SpecCycle+ MemWINV- VGASnoop- ParErr-
Stepping- SERR- FastB2B-
        Status: Cap- 66Mhz+ UDF- FastB2B+ ParErr- DEVSEL=fast >TAbort- <TAbort-
<MAbort- >SERR- <PERR-
        Latency: 0

00:01.1 SMBus: nVidia Corporation nForce3 SMBus (rev a4)
        Subsystem: Giga-byte Technology: Unknown device 0c11
        Control: I/O+ Mem- BusMaster- SpecCycle- MemWINV- VGASnoop- ParErr-
Stepping- SERR- FastB2B-
        Status: Cap+ 66Mhz+ UDF- FastB2B+ ParErr- DEVSEL=fast >TAbort- <TAbort-
<MAbort- >SERR- <PERR-
        Interrupt: pin A routed to IRQ 3
        Region 4: I/O ports at 1c00 [size=64]
        Region 5: I/O ports at 2000 [size=64]
        Capabilities: [44] Power Management version 2
                Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA
PME(D0-,D1-,D2-,D3hot+,D3cold+)
                Status: D0 PME-Enable- DSel=0 DScale=0 PME-

00:02.0 USB Controller: nVidia Corporation nForce3 USB 1.1 (rev a5) (prog-if 10
[OHCI])
        Subsystem: Giga-byte Technology: Unknown device 5004
        Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop-...

Ooops. Sorry, I meant to attach that, not paste it in.

Created an attachment (id=1917)
LSPCI -vv output

I was having exactly the same problems, AMD64 ATI 9200, random crashes in
firefox etc.

Solution was to install fglrx drivers and change xorg.conf to point to fglrx
instead of ati

driver:"ati"

look at

http://www.ubuntuforums.org/showthread.php?t=22496&highlight=ATi+crash
http://www.ubuntuforums.org/showthread.php?t=20191&highlight=amd64+fglrx for
more info

one note - the fglrx driver wont output via DVI, so you will have to plug your
monitor into the analogue output of your video card to get a display
http://www.ubuntuforums.org/showthread.php?t=21927&highlight=fglrx+blank+screen

For me the random hangs happen regardless of whether I use fglrx or not.

For grins I ran the Memtest 1.50+ tests 8 times. (Memtest ran for over 4 hours)
and found zero errors...

(In reply to comment #31)
> For grins I ran the Memtest 1.50+ tests 8 times. (Memtest ran for over 4 hours)
> and found zero errors...

I've noticed that you have 1 GB of ram. I assume that this is 2 x 512 MB double
sided sticks running at DDR400 or more?
According to AMD documents, this configuration (4 banks of ram @ DDR400) is not
supported by your CPU (on A64s the memory comtroller is on the CPU)! I have an
Athlon64 3000+ with 1 GB of ram and I was getting strange (but not often)
crashes for more than a year.
Prime95 would get an error within 1 minute. However, Memtest86+ ran fine for hours.

Recently I replaced on of my sticks with a single sided stick (hence only 3
banks of ram) and my problems have disappeared. No more random crashes and
Prime95 can run for hours.

Try pulling out one of your ram sticks and see what happens.
I am really annoyed at AMD and the mobo manufacturers for keeping this important
limitation so quiet. My motherboard manual says nothing about this.

(In reply to comment #32)
> I've noticed that you have 1 GB of ram. I assume that this is 2 x 512 MB double
> sided sticks running at DDR400 or more?

No. DDR333, though according to my motherboard manufacturer they support both
DDR400 and DDR333.

> Try pulling out one of your ram sticks and see what happens.

No offense, howevery every other operating system except for Ubuntu works. I'm
not pulling a memory stick for one out of seven different operating systems,
just because Ubuntu doesn't work right :)

In fact, I'm typing this right now in RHEL 3 WS, which runs fine for hours and
hours, unlike Ubuntu.

I'm experiencing the same behaviour (totally random hangs). However, my system
is pretty different. Please see attached lspci -vv output.

Created an attachment (id=2541)
lspci -vv output

Created an attachment (id=2542)
dmesg output

Created an attachment (id=2543)
Xorg.0.log

Created an attachment (id=2544)
xorg.conf

The hangs occur totally randomly, I don't see any pattern in them. They happen
with my current nVidia card as well as with the ATI Radeon 9000 I had before.
They don't occur with Fedora Core 3. If I'm watching TV or listening to music
when the hang occurs, the sound keeps playing, but the computer won't respond to
anything put a hard reset.

I have the same problems on my system. AMD64 2800+ MSI K8N Neo Plat with the
nForce 3 chipset. 768megs of Ram and an Asus ATI 9600XT. SB Audigy. The logs
never report a thing when it hangs. The screen freezes and the keyboard becomes
unresponsive... can't toggle the lights on any of the "lock" keys.

I haven't tried the Mem test yet. But I have two single sided mem modules in
there. one 256 and one 512. The system works fine in windows (unfortunately).

I tried pci=noacpi and that got me a hang at the login screen with the drum beat
sound repeating into eternity.

I too have tried the ati drivers and the fglrx drivers. No effect. I have a
dell 2005fpw so I'm running at 1680x1050... when I'm running.

I'm really intersted in getting this running. I have Ubuntu running on four
other systems with out any problems... I just want to run it on my power house
box now and it's not happening.

Rev

I followed the example in bug 16961
(http://bugzilla.ubuntu.com/show_bug.cgi?id=10668) and so far it seems to have
solved the problem.

If it all goes wrong again I'll post back here... but powernowd seems to have
been the solution.

Never mind... still hangs. =(

I tried updating to Ubuntu's kernel 2.6.11-k7

I completely removed powernowd from my system and rebooted.

Still hit a random lock not much later.

However, I think I found something.

One of the times it hardlocked, it suddenly started spewing:

gam_server (or the like) and then said something to the effect of 'kernel panic
- Aieeeee - killing interrupt handler'.

Fedora Core 3, Solaris 10, and other operating systems continue to run without
incident, only Ubuntu continues to have this issue.

do not mangle the bug severity. Also people, don't use 2.6.11. It's not supported.

One test that it seems nobody did was to try to disable DRI from X.org. That
would exclude
the only real kernel<->xorg interaction.

Fabio

(In reply to comment #44)
> do not mangle the bug severity. Also people, don't use 2.6.11. It's not supported.

I've tried 2.6.10-5, 2.6.11, nothing seems to make any real difference.

> One test that it seems nobody did was to try to disable DRI from X.org. That
> would exclude
> the only real kernel<->xorg interaction.

Well, considering DRI support isn't available for the Radeon 9800 or my new
video card, the GeForce 6800 GT I think you can safely assume the obvious.
Before you ask, this is with the *standard* "nv" and "radeon" drivers that ship
with Ubuntu. Not with "evil" proprietary binary drivers.

(In reply to comment #45)
> Not with "evil" proprietary binary drivers.

I must note I have tried the proprietary drivers as well just to see if it was
the standard Xorg drivers that were bad.

My personal gut feeling is that this is a kernel patch that Ubuntu has applied.

(In reply to comment #45)
> (In reply to comment #44)
> > do not mangle the bug severity. Also people, don't use 2.6.11. It's not
supported.
>
> I've tried 2.6.10-5, 2.6.11, nothing seems to make any real difference.
>
> > One test that it seems nobody did was to try to disable DRI from X.org. That
> > would exclude
> > the only real kernel<->xorg interaction.
>
> Well, considering DRI support isn't available for the Radeon 9800 or my new
> video card, the GeForce 6800 GT I think you can safely assume the obvious.
> Before you ask, this is with the *standard* "nv" and "radeon" drivers that ship
> with Ubuntu. Not with "evil" proprietary binary drivers.

That's why it would be interesting to see if force disabling it in Xorg solves
the problem.

(In reply to comment #44)
> do not mangle the bug severity.

Sorry about that, but random hangs seemed pretty critical to me and I was trying
to get more people to take interest in this bug since noone seems to know what's
going on. I think Ubuntu is the best distro ever, but in the absence of a fix or
a workaround to this bug I'll switch to FC4 as soon as it's out.

Ok let's do another test, i am not sure who is using what in here, but
if anybody is running breezy (and please do not update only for a test!)
can you try using 2.6.12??

Note: it's not 2.6.12 final! it's 2.6.12 RC5 and it's not to be considered safe
in production (yet)

(In reply to comment #44)
> do not mangle the bug severity. Also people, don't use 2.6.11. It's not supported.
>
> One test that it seems nobody did was to try to disable DRI from X.org. That
> would exclude
> the only real kernel<->xorg interaction.

DRI is not loading at all on my system, it's commented out and the X.org log shows nothing for it. I don't
know how to disable it more than that. I still have the hangs.

I'm getting this problem as well with an AMD 64 2800+ on a Via K8M800 with a
Radeon 7000 Series. So far it has happened in Firefox as well as in the
Synaptic Package Manager. All the lock ups have been within 5 minutes (usually
less) of booting up.

Ok, I'd like to get this fixed for breezy. I'm not sure if it still happens, so
someone please refresh the this bug.

What I need is for someone to force this bug to occur while in text console.
(use elinks, wget, or whatever to do some network traffic, since it seems to be
related to network).

When the crash occurs, please either write down, or take a picture of the crash
on the screen (all of it is important, not just the initial message).

The one person that did see something mentioned killing an interrupt handler. I
suspect it is the network card, but the mention of cam_server means it may also
be sound related.

I've seen enough "me too's" to bump this to "major"

This might be due to the dynamicClocks bugs being around.

Both the radeonfb and the xorg radeon driver contains bugs about the
dynamicClock powersaving feature.
These bugs are fixed in xorg cvs (6.8.3) and kernel (2.6.13).

kernel bug with little info: http://bugme.osdl.org/show_bug.cgi?id=4295
xorg bugs among others: https://bugs.freedesktop.org/show_bug.cgi?id=1912

no, dynamicclocks is disabled on non-mobility hardware.

This bug has been flagged because it is old and possibly inactive. It may or may
not be fixed in the latest release (Breezy Badger 5.10). It is being marked as
"NEEDSINFO". In two weeks time, if the bug is not updated back to "NEW" and
validated against Breezy, it will be closed.

This is needed in order to help manage the current bug list for the kernel. We
would like to fix all bugs, but need users to test and help with debugging.

If this change was in error for this bug, please respond and make the
appropriate change (or email <email address hidden> if you cannot make the
change).

Thanks for your help.

Can anyone please test latest breezy, or better, latest dapper kernel (or wait
for dapper flight 2 cd's)?

Hi,

is there an update on this, I have the exact same problem with my Kubuntu (Breezy), latest 686 kernel from repositorie.

I tried 386, 686 and k7 kernel and it does exactly the same, random crash of X.

The difference with Shawn is that I have a Nvidia 6600GT, I tried with "nv" and "nvidia" drivers same problem, even with the drivers from Nvidia's website.

My configuration is:
AMD Athlon64 3000+ (venice)
2GB of memory
Asus A8V Deluxe
120GB Seagate HDD
Nvidia 6600GT AGP 128Mo

I used Automatix (5.4.3 I think) to update some of the software.

Also, my machine runs a OpenSsh server and when the server hangs, it becomes inaccessible...

Hope it can help.

I might try to compile a more recent kernel from kernel,org, I'll keep you posted.

And I also confirm that this bug only happens with Ubuntu.

Ok, I compiled a new kernel yesterday 2.6.15 + archck4.1 (brings it up to 2.6.15.4 + speed optimisations), and it just happenned again, freeze while I was browsing the web with Firefox. I guess it's not the kernel then...

Oh I forgot to say. this time, only X was frozen, I could connect from my other computer with SSH and kill it.

After testing Dapper, the same problems happens, so I did some more tests and I finally found the source of the problem, it's in the AMD64_AGP modules, after blacklisting it, I had no more hangs.

Shawn, can you test the latest kernel in Dapper? If this is still an issue, please reopen this bugs. Thanks.

I don't know for Shun but for me disabling Cool'n' quiet sorted the problem out.

I tested Dapper kernel up to 20 I think.

I can't do more tests, I'm not using Ubuntu as my main Linux distrib anymore. I might reinstall it with the release.

Hi Corey,

I have retested using the latest Dapper drake beta. No lockups, and everything seems to work "just peachy." I'd say this has definitely been fixed. Thanks.

Hi to all.

I have the same bug on my Gentoo:
1) any video card: nvidia mx440, at x800pro, ati x800xt, ati 9600, nvidia 5400fx and some very-very old agp cards (i don't remember exactly wich)
1.1) with any video drivers: ati, nvidia, binary on not, or without any drivers (just xorg vesa driver)
2) any 64-bit gentoo-patched kernels since 2.6.8 up to now - 2.6.28
3) hangs appears most frequently in any browsers, while watching macromedia Flash, but also while running clean X-server.
4) in fluxbox less hangs, than in KDE.
5) in console (framebuffer on not) system had no hangs for one week up-time.
6) no hangs in 32-bit distros (free/net-bsd, gentoo, knoppix, mandriva, redhat, windows xp, vista)

PS sorry for my English.
PPS if any ideas, please mailto NazarovYV (symbol) mail.ru