[rv635] blank screen and hangs radeon 3650 on amd64

Bug #254583 reported by sam tygier
6
Affects Status Importance Assigned to Milestone
xserver-xorg-driver-ati
Fix Released
Medium
xserver-xorg-video-ati (Ubuntu)
Fix Released
Undecided
Unassigned

Bug Description

Tested with intrepid alpha 3 and 20080804 live cds on amd64.

default boot option gives usplash with progress bar. the bar completes, the screen goes black. unable to switch VTs.

with quiet removed, and nosplash set. standard stuff scrolls past. looks like it is going ok. then the screen flicks as if it is starting X and GDM, but just get blank black screen. unable to switch VTs.

booting with safe graphics mode. boots fine.

My graphics card is a PCI express radeon HD 3650 (ASUS EAH3650 SILENT)
08:00.0 VGA compatible controller: ATI Technologies Inc Mobilitiy Radeon HD 3600 Series

Revision history for this message
sam tygier (samtygier) wrote :
Revision history for this message
sam tygier (samtygier) wrote :
Revision history for this message
sam tygier (samtygier) wrote :
Revision history for this message
sam tygier (samtygier) wrote :

if i set the xorg.conf to use VESA, GDM starts ok. if i set it to radeon then i get the freeze at a black screen

Revision history for this message
sam tygier (samtygier) wrote :

i have installed. i set the driver to radeon. (see attached xorg.conf).

now when X starts the whole computer reboots.

Revision history for this message
sam tygier (samtygier) wrote :

PS: there is no new log file write. or atleast it does not get flushed to disk before the reboot

Revision history for this message
sam tygier (samtygier) wrote :

Bug 252927 is about the same card, and is similar, grey vs black screen.

Revision history for this message
Carlos Nucette (cjnucette-gmail) wrote :

Sorry if this is not the place to post this, but i have the same problem but with an integrated intel video card (Intel Corporation Mobile GM965/GL960 Integrated Graphics Controller) on a Vostro 1500, first parts of the progress bar are all over the place and then it goes blank even booting whit safe graphics mode.

Revision history for this message
Carlos Nucette (cjnucette-gmail) wrote :

Ok this is weird, if i use the default language and keyboard it boots ok, but if i choose spanish and latinamerican keyboard it goes blank.

Revision history for this message
Tormod Volden (tormodvolden) wrote : Re: [rv635] boots to blank screen unless safe graphics mode enabled radeon 3650

Thanks for your report. Please take away "splash" from the boot parameters. Can you try to debug it remotely via ssh and get the log from the default configuration (no driver specified, should use "ati")?

Can you also please try the newest driver in intrepid and if it's still the same file a bug upstream on bugs.freedesktop.org (Product=xorg, Component=Driver/Radeon), attaching your Xorg.0.log?

Changed in xserver-xorg-video-ati:
assignee: nobody → tormodvolden
status: New → Incomplete
Revision history for this message
In , sam tygier (samtygier) wrote :

When I start the X server the screen goes blank and machine locks up hard. Caps key does not light caps LED. cant switch VT. can't ping. sometimes the machine reboots.

I see this with the radeon driver, vesa works fine.

I have a ASUS EAH3650 SILENT, ATI radeon HD 3650
08:00.0 VGA compatible controller: ATI Technologies Inc Mobilitiy Radeon HD 3600 Series

I am running ubuntu intrepid (dev branch). with xserver-xorg-core 2:1.4.99.906-1ubuntu3 and xserver-xorg-video-radeon 1:6.9.0+git20080802.1f3eee36-1ubuntu1

the bug is reported downstream at
https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/254583

you can find some strace and gdb logs there. I can copy them here if that is useful for you.

Revision history for this message
In , Bugzi09-fdo-tormod (bugzi09-fdo-tormod) wrote :

From the gdb debugging, it can be seen that it dies (hangs) in
RADEONMapMMIO -> pci_device_map_range -> mmap
See the gdb log http://launchpadlibrarian.net/16676661/xorg.3650.gdb4.log from the Ubuntu bug for more details, function arguments, local variables etc.

Revision history for this message
sam tygier (samtygier) wrote : Re: [rv635] boots to blank screen unless safe graphics mode enabled radeon 3650

i got the radeon update as of this morning.
i removed splash quiet and xforcevesa from the grub line (boot showed oldskool boot messages)
i commented out Driver from my xorg.conf
login over ssh
started GDM (i have remove the symlinks so that GDM did not start at boot)
screen goes blank. ssh session is unresponsive. wont respond to pings. caps lock key does not light caps LED.

when i reboot the Xorg.0.log is from earlier today when i started GDM with the vesa driver. so no log is written.

any ideas for further debugging?

Revision history for this message
In , Corbin Simpson (mostawesomedude) wrote :

Just offhand, sounds like fglrx is present in the kernel, blocking a successful MMIO setup.

Were you using fglrx before?

Revision history for this message
In , sam tygier (samtygier) wrote :

no. i had the linux restricted modules installed, but never the fglrx packages (not on this install anyway).

i have removed the restricted-modules, rebooted, and the problem persists.

Revision history for this message
Tormod Volden (tormodvolden) wrote :

Please try the device Option "DRI" "off". I would also suggest you try to start X using strace to see approximately how far it gets before it hangs:
 sudo strace Xorg
or
 sudo strace -o xorg.strace.log Xorg

You can also try to start it with gdb, setting breakpoints to narrow it down.

Revision history for this message
sam tygier (samtygier) wrote :

with the boot options as before, and DRI off, i get the same lockup.
(just to note the magic sysrq keys don't work)

when i reboot after straceing to a file, the file is not there (and its not in lost and found).

strace in an ssh session gave some output :-)

Revision history for this message
sam tygier (samtygier) wrote :

seemed to get a few lines further on my second attempt. maybe useful.

i dont know much about setting break points. you would have to hold my hand. am ssam on IRC

Revision history for this message
sam tygier (samtygier) wrote :

sorry, i pasted wrong. the last upload can be ignored/deleted.
the bonus new lines are:

write(0, "(II) Setting vga for screen 0.\n", 31) = 31
write(0, "(II) RADEON(0): MMIO registers a"..., 64) = 64
open("/sys/bus/pci/devices/0000:08:00.0/resource2", O_RDWR) = 9
mmap(NULL, 65536, PROT_READ|PROT_WRITE, MAP_SHARED, 9, 0

Revision history for this message
Tormod Volden (tormodvolden) wrote :

Thanks, I think that can help the upstream developers already. Can you please also attach the output from "lspci -vvnn"?

Revision history for this message
Tormod Volden (tormodvolden) wrote :

If you want to play with gdb, the above message is probably from RADEONMapMMIO() in src/radeon_driver. Maybe the call to xf86MapPciMem(). You can for instance set a breakpoint at the beginning of this function:
(gdb) breakpoint RADEONMapMMIO
(gdb) run
and then when it stops, use "next" to step through the lines.
(gdb) next

See also the gdb links on https://wiki.ubuntu.com/X/Backtracing

Revision history for this message
sam tygier (samtygier) wrote :

i think that needed to be "break RADEONMapMMIO"
ok. i nexted until it died. log attached.

I also tried booting with mem=512M and mem=1024M but this gave the same results.

also the slowness of GDB seemed to give it time to write the Xorg.0.log to disk. so i will attach that.

Revision history for this message
sam tygier (samtygier) wrote :
Changed in xserver-xorg-driver-ati:
status: Unknown → Confirmed
Revision history for this message
sam tygier (samtygier) wrote :

some further gdbing. thanks to tormod for the help

Revision history for this message
sam tygier (samtygier) wrote :

got even further this time.

Changed in xserver-xorg-video-ati:
assignee: tormodvolden → nobody
status: Incomplete → Confirmed
Revision history for this message
In , sam tygier (samtygier) wrote :

Just to confirm, the only files on the system with fglrx in the name, are the install scripts in ubuntu's jockey, and the .desktop file in app-install.

I did a cold start (machine had been off for hours). booted with mem=1024M, nosplash screen. the machine still locked and rebooted when starting the X server.

Revision history for this message
sam tygier (samtygier) wrote :

I tried the 32bit alpha 4 cd. and it is broken in a different way. It seems to boot fine, but the screen is all white. compiz seems to be running, because if I do Super+E, i get the zoom out to see multiple desktops effect (zoom out to see 2 white squares).

Revision history for this message
Tormod Volden (tormodvolden) wrote :

That's sounds like a compiz bug. Does everything else work fine without compiz? Log in using a "failsafe gnome" session, or disable compiz with:
 touch $HOME/.config/compiz/disable-compiz

Revision history for this message
In , sam tygier (samtygier) wrote :

Created an attachment (id=18566)
fixed Xorg.0.log

Revision history for this message
In , sam tygier (samtygier) wrote :

this is fixed with latest ubuntu updates 1:7.4~1ubuntu1.
X.Org X Server 1.4.99.906 (1.5.0 RC 6)

Bryce Harrington (bryce)
Changed in xserver-xorg-video-ati:
status: Confirmed → Triaged
Revision history for this message
sam tygier (samtygier) wrote :

This fixed with the 20080829 live cd. either the new kernel, or the xorg updates must have sorted it. thanks

Changed in xserver-xorg-video-ati:
status: Triaged → Fix Released
Revision history for this message
sam tygier (samtygier) wrote :
Changed in xserver-xorg-driver-ati:
status: Confirmed → Fix Released
Revision history for this message
In , sam tygier (samtygier) wrote :

It seems that it was the kernel upgrade from 2.6.26 to 2.6.27 that actually fixed this.

Revision history for this message
In , Robin H. Johnson (robbat2) wrote :

I ran into this in 2.6.27-rc5-00006-gbef69ea yesterday (latest Git kernel), and latest Git ati driver, so I don't think it's fixed in new kernels.

Testing specifics:
- To trigger the bug, you MUST cause the card to POST with a display on the DVI-0 output. If there is only a display on the DVI-1 output, it works perfectly. Simply disconnecting the DVI-0 output is not sufficent.

I did however trace it using poor-mans debugging (print statements). In src/radeon_modes.c, the RADEONProbeOutputModes, this block:
285 if (output->MonInfo)
286 modes = xf86OutputGetEDIDModes (output);

It's inside that call that it's hanging. After commenting out those two lines, X comes up perfectly for me.

Maybe related, maybe unrelated, when I exit X with that change, sometimes the contents of DVI-0 end up on both monitors again, frozen so the console isn't restored.

Revision history for this message
sam tygier (samtygier) wrote :

further investigation (booting with the 2.6.26-5-generic kernel), reveals that this was a kernel issue.
moving to 2.6.27 seems to have been the fix

Revision history for this message
Victor Van Hee (vvanhee) wrote :

I haven't been able to get this to work (Radeon 3650) with Alpha 5 64 bit live cd, or even if I install the full intrepid distro and then update it. Is the newer kernel part of the typical upgrade process, or should I wait for the next alpha (or beta)?

See https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/264462 for further details.

Revision history for this message
Mnemonic (mnemonic-get2net) wrote :

I have the same Card ATI Mobility Radeon HD 3650 and the same error. I tryed using Ubuntu 8.10 and 9.04 fully updated but nothing works. What was the fix?

Revision history for this message
Géza Búza (medve) wrote :

I can also confirm this issue. I have Ubuntu / jaunty on lenovo t500. I tried to intall the latest (2009-07-23) driver from ati.amd.com (ati-driver-installer-9-7-x86.x86_64.run), but I got blank screen again.

Changed in xserver-xorg-driver-ati:
importance: Unknown → Medium
Changed in xserver-xorg-driver-ati:
importance: Medium → Unknown
Changed in xserver-xorg-driver-ati:
importance: Unknown → Medium
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.