X fails to start after recent update

[D. Hugh Redelmeier]

I did an update.  It said I should reboot.
After reboot, X does not come up.

I can flip to a text console:  that's what I'm typing this into.

Anyone else have this problem?

I'm running X86-66 Fedora 10.  The video card is an Asus EAH3650
Silent Magic.

/var/log/Xorg.0.log shows some bad news:

(EE) RADEON(0): Output DVI-0 enabled but has no modes

But I don't know why it says that.  Preceding lines include:
  (II) RADEON(0): Printing probed modes for output DVI-0
  (II) RADEON(0): Modeline "2560x1600"x59.9  268.00  2560 2608 2640 2720  1600 1603 1609 1646 +hsync +vsync (98.5 kHz)
This is a correct modeline for the monitor
  (II) RADEON(0): Output HDMI-0 disconnected
  (II) RADEON(0): Output DVI-0 connected
  (II) RADEON(0): Using exact sizes for initial modes

Eventually, it says:
  (EE) Screen(s) found, but none have a usable configuration.

Fatal server error:
  no screens found

I don't know which update is the problem.  Certainly one in the alst
few days.

It cannot be the actual X device driver because I didn't update that.
There is a driver update but I don't apply it because it does not
work on my system.  See
  https://bugzilla.redhat.com/show_bug.cgi?id=484636

The ati driver recognizes the video card as:
  (--) RADEON(0): Chipset: "ATI ATI Radeon HD 3600 XT" (ChipID = 0x9598)
The "ATI ATI" is a typo fixed in the current driver.

[Ranbir]

On Thu, 2009-02-26 at 16:57 -0500, D. Hugh Redelmeier wrote:
> I did an update.  It said I should reboot.
> After reboot, X does not come up.
>
> I can flip to a text console:  that's what I'm typing this into.
>
> Anyone else have this problem?

One of the recent Xorg and/or fglrx (from rpmfusion) updates has borked
my 3D: it no longer works.  I logged out, and when I logged back in,
that's when I realized compiz wouldn't start.  I've spent an hour trying
to fix it, but I'm just bumbling around.

These are the errors I'm seeing in the Xorg log:

(EE) AIGLX error: fglrx exports no extensions
(/usr/lib64/dri/fglrx_dri.so: undefined symbol: __driDriverExtensions)
(EE) AIGLX: reverting to software rendering

I have an ATI HD3850.

What's happened? How do I fix it?  I wish the 3D bits weren't so touchy.

Regards,

Ranbir

[Wolfgang S. Rupprecht]

"D. Hugh Redelmeier"  writes:
> I don't know which update is the problem.  Certainly one in the alst
> few days.

Try reverting the Xorg server back one notch.  That is what worked for
me.

rpm -Uvh --oldpackage xorg-x11-server-Xorg-1.5.3-6.fc10.x86_64.rpm

(I have keepcache=1 in /etc/yum.conf so the old rpm was sitting around
in /var/cache/yum/updates/packages/ .)

-wolfgang

[Andrew Kelly]

Hugh,

On Thu, 2009-02-26 at 16:57 -0500, D. Hugh Redelmeier wrote:
> I did an update.  It said I should reboot.
> After reboot, X does not come up.
>
> I can flip to a text console:  that's what I'm typing this into.
>
> Anyone else have this problem?

Same problem here, also running F 10 on X86-66 and with an Asus EAH3650
Silent Magic. Although, when my stuff hangs, it's hung, I can't get to a
text console at all. I have to do a hard reboot (cringe) and then grab
grub on the way up to boot into run level 3.

I can't tell you exactly what's choking and dying, BUT, I have at least
been able to narrow it down the exact rpm that is doing the damage.
It's xorg-x11-drv-ati-6.10.0-2.fc10.i386.rpm

It's probably either the ati or radeo driver modules, but that's just
speculation at this point.

My work-around was kind of brain dead, but it's kept me running. I
queried the rpm for its files, then tar'ed myself an archive of those
before updating, and played them back after the update. 
Really, really cheesy, but I've got 9 irons in the fire right now and
can't look back at the moment
> I'm running X86-66 Fedora 10.  The video card is an Asus EAH3650
> Silent Magic.
>
> /var/log/Xorg.0.log shows some bad news:
>
> (EE) RADEON(0): Output DVI-0 enabled but has no modes
>
> But I don't know why it says that.  Preceding lines include:
>   (II) RADEON(0): Printing probed modes for output DVI-0
>   (II) RADEON(0): Modeline "2560x1600"x59.9  268.00  2560 2608 2640 2720  1600 1603 1609 1646 +hsync +vsync (98.5 kHz)
> This is a correct modeline for the monitor
>   (II) RADEON(0): Output HDMI-0 disconnected
>   (II) RADEON(0): Output DVI-0 connected
>   (II) RADEON(0): Using exact sizes for initial modes
>
> Eventually, it says:
>   (EE) Screen(s) found, but none have a usable configuration.
>
> Fatal server error:
>   no screens found
>
> I don't know which update is the problem.  Certainly one in the alst
> few days.
>
> It cannot be the actual X device driver because I didn't update that.
> There is a driver update but I don't apply it because it does not
> work on my system.  See
>   https://bugzilla.redhat.com/show_bug.cgi?id=484636
>
> The ati driver recognizes the video card as:
>   (--) RADEON(0): Chipset: "ATI ATI Radeon HD 3600 XT" (ChipID = 0x9598)
> The "ATI ATI" is a typo fixed in the current driver.

Andy

[D. Hugh Redelmeier]

| From: Andrew Kelly

| Same problem here, also running F 10 on X86-66 and with an Asus EAH3650
| Silent Magic. Although, when my stuff hangs, it's hung, I can't get to a
| text console at all. I have to do a hard reboot (cringe) and then grab
| grub on the way up to boot into run level 3.
| 
| I can't tell you exactly what's choking and dying, BUT, I have at least
| been able to narrow it down the exact rpm that is doing the damage.
| It's xorg-x11-drv-ati-6.10.0-2.fc10.i386.rpm

Actually, you are hitting a different problem.  One I hit earlier.  I
mentioned this one in my message:

| > There is a driver update but I don't apply it because it does not
| > work on my system.  See
| >   https://bugzilla.redhat.com/show_bug.cgi?id=484636

Please add anything you can to the bugzilla entry.  I presume that it
will be taken more seriously if more than one person reports
experiencing it.

| My work-around was kind of brain dead, but it's kept me running. I
| queried the rpm for its files, then tar'ed myself an archive of those
| before updating, and played them back after the update. 
| Really, really cheesy, but I've got 9 irons in the fire right now and
| can't look back at the moment

That isn't the best way to work around the problem.  By downgrading
the .rpm, you are still in known territory.  And you aren't lying to
the package manager.

You can fetch old versions of each package from koji.  Here's the page
for this package:
  http://koji.fedoraproject.org/koji/packageinfo?packageID=95

[D. Hugh Redelmeier]

| From: Wolfgang S. Rupprecht

| "D. Hugh Redelmeier"  writes:
| > I don't know which update is the problem.  Certainly one in the alst
| > few days.
| 
| Try reverting the Xorg server back one notch.  That is what worked for
| me.
| 
|     rpm -Uvh --oldpackage xorg-x11-server-Xorg-1.5.3-6.fc10.x86_64.rpm
| 
| (I have keepcache=1 in /etc/yum.conf so the old rpm was sitting around
| in /var/cache/yum/updates/packages/ .)

Were your symptoms like mine or did you hit a different problem?

Have you found or created a bugzilla entry for your problem?

I hope to get the time to do the bugzilla thing (i.e. add to a
previous report or create a new one) in the next couple of days.

I found that I could get my system up and running by reverting two
packages.  I guessed that they depend on each other so I reverted
both:
  xorg-x11-server-common-1.5.3-13.fc10.x86_64
  xorg-x11-server-Xorg-1.5.3-13.fc10.x86_64
The previous version was 1.5.3-6.fc10.

Do you know if you can mix and match different levels of
xorg-x11-server-common and xorg-x11-server-Xorg?
server-Xorg requires server-common, and not the reverse, but I don't
know how to query about version constraints.

I fetched the old versions from koji.  Handy facility:
  http://koji.fedoraproject.org/koji/buildinfo?buildID=74390

You can see the change history from the new version's koji page:
  http://koji.fedoraproject.org/koji/buildinfo?buildID=82838
If I had to guess which change caused the problem, I'd pick this one:

* Mon Dec 22 2008 Adam Jackson  1.5.3-8
    - xserver-1.5.3-aspect-me-harder.patch: Fix mode selection when only one
      monitor is present and it doesn't claim a preferred mode.
    - xserver-1.5.0-bad-fbdev-thats-mine.patch: Do the same for sbus that we
      do for pci.
    - xserver-1.5.0-hide-cursor.patch: Backport fixes to unhide logic from master.

[Wolfgang S. Rupprecht]

"D. Hugh Redelmeier"  writes:
> Were your symptoms like mine or did you hit a different problem?

Hard to say.  My screen started out displaying the boot status for the
daemons (I commented out rhgb in grub.conf).  So my screen wasn't blank,
it showed the last thing that the boot sequence displayed.

The messages in Xorg.0.log were similar with no usable modes found and
the Xorg program giving up.  In my case it segv-ed, but that may be an
insignificant difference (perhaps some uninitialised mode-line data was
more benign in your case.)

I find it interesting that we effectively had the same monitor (yours
was a Dell and mine a Samsung, but otherwise had the same resolution and
modes.)
> Have you found or created a bugzilla entry for your problem?

Yes.   https://bugzilla.redhat.com/show_bug.cgi?id=487285
> Do you know if you can mix and match different levels of
> xorg-x11-server-common and xorg-x11-server-Xorg?
> server-Xorg requires server-common, and not the reverse, but I don't
> know how to query about version constraints.

In general I'd not mix and match.  In this case I "winged it" because
the major and minor Xorg versions were the smea and I figured the
ABI/API didn't change.
> If I had to guess which change caused the problem, I'd pick this one:
> * Mon Dec 22 2008 Adam Jackson <[e-mail]> 1.5.3-8
> - xserver-1.5.3-aspect-me-harder.patch: Fix mode selection when only one
>   monitor is present and it doesn't claim a preferred mode.
> - xserver-1.5.0-bad-fbdev-thats-mine.patch: Do the same for sbus that we
>   do for pci.
> - xserver-1.5.0-hide-cursor.patch: Backport fixes to unhide logic from master.

I agree.  That one looks darn close to what we are seeing.

-wolfgang