[RV710 & RV730] X freezes starting on 10th April

Just thought I should add this might be better described as a hang or freeze, rather than a crash.

Same thing for me on an HD 4650.

Again on a HD 4650

I've the same problem with Radeon Driver and ATI HD4670 .... random freeze with first black screen and after white /light grey screen .. and necessary to do a hard reboot. ;)

Excuse my poor english ;)

Same issue here with Lucid Lynx on HP dv6, ATI Mobility Radeon HD4530 and free driver.

And here, Acer E360, ATI HD4350 pcie, 32bit lucid (fully up to date as of 20 minutes ago).

I'll try and gather logs next time it happens.

I noticed the following entries in the boot dmesg messages:

[ 17.846987] [drm] radeon defaulting to kernel modesetting.
[ 17.846990] [drm] radeon kernel modesetting enabled.
[ 17.847037] radeon 0000:01:00.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[ 17.847042] radeon 0000:01:00.0: setting latency timer to 64
[ 17.849248] [drm] radeon: Initializing kernel modesetting.
[ 17.863917] [drm] register mmio base: 0xDFDF0000
[ 17.863920] [drm] register mmio size: 65536
[ 17.863978] ATOM BIOS: 113
[ 17.863993] [drm] Clocks initialized !
[ 17.864016] mtrr: type mismatch for c0000000,10000000 old: write-back new: write-combining
[ 17.864019] [drm] Detected VRAM RAM=256M, BAR=256M
[ 17.864020] [drm] RAM width 64bits DDR
[ 17.869406] [TTM] Zone kernel: Available graphics memory: 1996288 kiB.
[ 17.869422] [drm] radeon: 256M of VRAM memory ready
[ 17.869424] [drm] radeon: 512M of GTT memory ready.
[ 17.869462] alloc irq_desc for 29 on node -1
[ 17.869464] alloc kstat_irqs on node -1
[ 17.869473] radeon 0000:01:00.0: irq 29 for MSI/MSI-X
[ 17.869478] [drm] radeon: using MSI.
[ 17.869506] [drm] radeon: irq initialized.

Perhaps booting with radeon.modeset=0 fixes this; or booting with pci=nomsi ( https://bugs.launchpad.net/ubuntu/+source/linux/+bug/509273 )

I can confirm the description in the original report. I can also confirm that the problem has appeared around the day 10th April.

It's been suggested this may be a duplicate of Bug #557829 however I think this bug appeared only after that one was reported.

People might also be interested in the 2nd post of: http://ubuntuforums.org/showthread.php?s=8674fdc2aaebd69745b39dd9f877ed8b&p=9146088 It suggests this bug may cause hardware damage and strongly suggests switching to fglrx in the meantime.

Same problem here on ATI Radeon HD 4600 Series Device ID 9490. Proprietary driver is not used.

I can reproduce this bug by switching to open source driver from fglrx.
ATI Radeon HD4670 here.

Please stop posting 'me-too's if you have nothing more than that to say, just tick the 'affects me too' button. 'me-toos' just bloat the bug report and make it harder for developers to read the report.

If it is correct that this regression started April 10th, that roughly coincides to when we brought in the 6.13.0 version of -ati.

It would be helpful of those of you able to reproduce this problem could downgrade to version 1:6.12.192-2ubuntu2 and verify the issue goes away. If it does not go away, then the bug is not likely to be in xserver-xorg-video-ati. After that, the next thing to check would be to boot an earlier kernel from before April 10th and see if that makes the issue go away.

In either case, I think this bug report should be sent upstream, as the changes we've added to ubuntu are almost entirely cherrypicks from upstream, so this is probably relevant to them.

Willing to downgrade to 1:6.12.192-2ubuntu2 to need guidance on how do do that.

Happens to me on ATI Mobility Radeon HD 4570

Radeon HD 4850 here and no problems whatsoever.

I can confirm this on my Mobility Radeon HD 4570.

It happens only with current radeon module if 3D effects (Compiz) are enabled.
I had 5 days uptime with current radeon module (1:6.13.0-1ubuntu5) and NO Compiz enabled.
I had 4 days uptime with fglrx proprietary module and BOTH compiz enabled or disabled.

The system hangs up randomly. Just before a striped green screen appears, mouse freezes.
Hang-ups started after 6.13.* version of xserver-xorg-video-{radeon,ati} packages appeared in the repos.
Last working version (10+ days of uptime) was https://launchpad.net/ubuntu/+source/xserver-xorg-video-ati/1:6.12.192-2ubuntu2

I've made a downgrade package of this version for convenience of testers:

  https://edge.launchpad.net/~ubuntu-x-swat/+archive/x-retro

I'm really skeptical that this would fix it however. There's only a tiny number of changes between the two versions, none of which look like things which could cause a freeze, but who knows, these freezes can often be sneaky things.

I'd be more curious if the Apr 10th date coincided with an update to the kernel. Downgrading the kernel might be a more productive thing to test.

I'm on my way in testing your downgrade packages, thanks.

There was also a kernel update on that date, indeed.

Since i am getting the same hangs i will try those packages and ( separately ) downgrading the kernel. Let's hope it helps to nail down the reason.

Downgrading these packages didn't solve the issue. It just gained a few more hours of uptime than before, and this time screen got totally blank instead of "green-striped".

Downgrade the kernel then ? What was the version introduced on 10th ?

On Fri, Apr 23, 2010 at 09:26:03AM -0000, Stefanauss wrote:
> Downgrading these packages didn't solve the issue. It just gained a few
> more hours of uptime than before, and this time screen got totally blank
> instead of "green-striped".

Okay, that's what I suspected. It's probably just coincidental that it
stayed up longer. Maybe the stars were aligned differently that moment.

Resetting to complete while folks test kernels. Check your /var/log/dpkg.log to see what got upgraded when.

On 13/04 i upgraded simultaneosly

2010-04-13 17:52:07 upgrade xserver-xorg-video-radeon 1:6.13.0-1ubuntu1 1:6.13.0-1ubuntu2

and

2010-04-13 17:51:49 upgrade linux-headers-2.6.32-20 2.6.32-20.29 2.6.32-20.30

Regarding my comment in #7, I've been running for the last 4 days without problems after disabling KMS (radeon.modeset=0).

So I'm inclined to agree that the problem is down at the kernel level rather than in the driver itself.

Disabling KMS workarounded this issue for me.

I was wondering. Since this bug still affects many user with a *default* configuration, which is KMS and desktop effects with the radeon module ENABLED, wouldn'it be appropriate to change default settings? How is ubuntu policy about that?

On Mon, Apr 26, 2010 at 06:14:45PM -0000, Stefanauss wrote:
> Disabling KMS workarounded this issue for me.
>
> I was wondering. Since this bug still affects many user with a *default*
> configuration, which is KMS and desktop effects with the radeon module
> ENABLED, wouldn'it be appropriate to change default settings? How is
> ubuntu policy about that?

I dunno, we also saw the case where a bunch of people found KMS caused
freezes, so we turned it off by default on those chips, and subsequently
a different bunch of people with the same hardware turned up to say
turning it off caused freezes for them.

Who knows what other problems could exist in the non-KMS case; we've
been running with KMS enabled by default for most of the release, so
blindly switching it off at this point seems like playing Russian
Roulette.

I think the better approach is to understand exactly why the freeze
happens and fix it. After all, at some point there won't be an option
other than KMS so this bug really does have to be fixed.

Alright, well the recent comments at least seem pretty convincing that the issue is a kernel regression, so I'll go ahead and reassign this to the kernel team at this point.

Hopefully JFo and the kernel team can help pinpoint the source of this breakage more precisely.

Hi Greg,

If you could also please test the latest upstream kernel available that would be great. It will allow additional upstream developers to examine the issue. Refer to https://wiki.ubuntu.com/KernelMainlineBuilds . Once you've tested the upstream kernel, please remove the 'needs-upstream-testing' tag. This can be done by clicking on the yellow pencil icon next to the tag located at the bottom of the bug description and deleting the 'needs-upstream-testing' text. Please let us know your results.

Thanks in advance.

    [This is an automated message. Apologies if it has reached you inappropriately; please just reply to this message indicating so.]

I have been using the fglrx driver for the last few days. Have just switched back to 'normal' Lucid. If (and I imagine it will) I get a crash I will try the v2.6.34-rc5-lucid kernel I have just installed from http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.34-rc5-lucid/ as I believe I've been instructed above. Will report back again later.

Cannot confirm on my HD4670.

I haven't seen the problem recently. Not sure what has changed.

On 4 May 2010 13:20, HX_unbanned <email address hidden> wrote:
> Cannot confirm on my HD4670.
>
> --
> [RV710 & RV730] X freezes starting on 10th April
> https://bugs.launchpad.net/bugs/565790
> You received this bug notification because you are a direct subscriber
> of the bug.
>

I have reverted my set-up to radeon, kms-enabled, compiz-enabled. With the current kernel image (2.6.32.21.22) the problem just seems to have disappeared. My current uptime is 2 days.