Bug 708462 - Udev fails to load if graphics mode set to switchable
Summary: Udev fails to load if graphics mode set to switchable
Keywords:
Status: CLOSED WONTFIX
Alias: None
Product: Fedora
Classification: Fedora
Component: xorg-x11-drv-ati
Version: 16
Hardware: i386
OS: Linux
unspecified
high
Target Milestone: ---
Assignee: X/OpenGL Maintenance List
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard:
: 698511 (view as bug list)
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2011-05-27 17:58 UTC by Zach Tomkoski
Modified: 2013-02-14 02:30 UTC (History)
15 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2013-02-14 02:30:16 UTC
Type: ---


Attachments (Terms of Use)
radeon init fault on lenovo T500 switchable graphics enabled (10.77 KB, text/plain)
2011-06-24 09:31 UTC, Michael Lausch
no flags Details
dmesg output if only radeon card is enabled in BIOS (81.88 KB, text/plain)
2011-06-24 09:43 UTC, Michael Lausch
no flags Details
dmesg when booting with rdblacklist=radeon (121.43 KB, text/plain)
2011-06-25 10:05 UTC, Michael Lausch
no flags Details
dmesg when booting with rdblacklist=i915 (121.45 KB, text/plain)
2011-06-25 10:08 UTC, Michael Lausch
no flags Details

Description Zach Tomkoski 2011-05-27 17:58:36 UTC
Description of problem: 

I have a Thinkpad T500 with switchable graphics.  If the switchable graphics mode is enabled the kernel will crash while attempting to load.

Version-Release number of selected component (if applicable):


How reproducible:

Everytime

Steps to Reproduce:
1. Enter BIOS > Graphics
2. Set Graphics Mode to Switchable
3. Attempt to boot into Fedora 15
  
Actual results:
Kernel Crashes and computer never loads

Expected results:
Fedora boots regularly

Additional info:
The system boots fine if you set the graphics to either Discrete or Integrated.  Just setting it to Switchable causes it to crash. Did not occur in F-14

Comment 1 Harald Hoyer 2011-05-30 07:55:22 UTC
if the kernel crashes, it's not the fault of udev.

Comment 2 Mebus 2011-06-01 20:53:44 UTC
I have got the same problem on a Thinkpad T400.

My internal graphics chip is the usual intel one and the descrete chip is a 

ATI Technologies Inc Mobility Radeon HD 3400 Series

I addition the fan of the Notebook is very noisy if I boot it up with the descrete graphics chip activated in the BIOS.

The system boots with the old Fedora 14 kernel.

Comment 3 dakar 2011-06-08 22:16:13 UTC
I have got the same problem on my Thinkpad T400 with "Switchable" graphics enabled in the BIOS too.

If I try to boot with "intel" or "radeon" driver the boot fails (when launching the xserver i think). Booting with "vesa" (in xorg.conf) is working.
There is no splash screen while booting and after the boot hangs and you can still login by switching to "Ctrb+Alt+F2".

There is also no "/sys/kernel/debug/vgaswitcheroo/" directory like it should be,  since vga_switcheroo was added to the kernel 2.6.34.

Comment 4 Michael Lausch 2011-06-15 13:10:43 UTC
I have a Lenovo T500 laptop with Intel integrated graphics and Radeon discrete graphics. vgaswitcheroo works fine with Fedora 14, at least for the one time i tried it. Botting with the latest Fedora 15 kernel fails if the BIOS configures the graphics in "switchable" mode. It works fine if either the integrated Intel chip or the discrete Radeon chip is selected in the BIOS.

I can't catch the complete kernel crash message because the notebook does not have a serial ine and i didn't have the time to recompile a kernel with compiled in network and netconsole support, which i hope is initialized *before* the KMS magic start.

Comment 5 Dave Airlie 2011-06-23 08:05:57 UTC
can you try booting with nomodeset?, then ssh in,

rmmod i915 radeon ; modprobe i915 modeset=1 ; modprobe radeon modeset=1

you might get the dmesg then.

I'll see if I can grab a machine tomorrow to test it.

Comment 6 Michael Lausch 2011-06-23 13:30:56 UTC
Tried this two times:
try 1:
Booted with nomodeset kernel cmd line.
removed both the i915 and radeon modules
modprobe i915 modeset=1
works flawless, screen is switched to smaller font, can start X
modprobe radeon modeset=1
works, can power off the the radeon chipset using vga_switcheroo sys debug file


try2:
Booted with nomodeset kernel cmd line.
removed both the i915 and radeon modules
modprobe radeon modeset=1
screen still has the large font. pressing Return in the console does not change anything (i hoped the shell prompt scrolls into view). 
shows no errors in the dmesg output (logged in via ssh, because screen is unusable)

is suspect some unhealthy interaction with vga_switcheroo during boot. 
Maybe the kernel video drivers are initialized before vga_switcheroo. I'm prepared to do some debugging using netconsole, but have to wait until i'm home again, because there's no ethernet LAN where i am now.

Comment 7 Michael Lausch 2011-06-24 09:31:46 UTC
Created attachment 509712 [details]
radeon init fault on lenovo T500 switchable graphics enabled

Captured the boot kernel messages using netconsole. I guess the second fault is is a follow up error. I also saw a radeon ringbuffer overflow error
[    4.789369] radeon 0000:01:00.0: IH ring buffer overflow (0xFFFFFFFF, 0, 15)

Comment 8 Michael Lausch 2011-06-24 09:43:18 UTC
Created attachment 509719 [details]
dmesg output if only radeon card is enabled in BIOS

Enabled only the discrete graphics in the BIOS config section. Machine boots up with no problems. Attached the dmesg output of a successful boot.

Comment 9 Chuck Ebbert 2011-06-24 11:43:45 UTC
*** Bug 698511 has been marked as a duplicate of this bug. ***

Comment 10 Michael Lausch 2011-06-24 11:55:32 UTC
Seems like MSI interrupt allocation or handling does not work in this case and the driver overflows the ringbuffer.

Comment 11 Michael Lausch 2011-06-25 10:05:04 UTC
Created attachment 509889 [details]
dmesg when booting with rdblacklist=radeon

Starting kernel with rdblacklist=radeon.
Modeswiching works, vgaswitcheroo works and let me poweroff the radeon card.

# cat /sys/kernel/debug/vgaswitcheroo/switch
0:DIS: :Off:0000:01:00.0
1:IGD:+:Pwr:0000:00:02.0

I added some debugging printk. Don't be confused.

Comment 12 Michael Lausch 2011-06-25 10:08:06 UTC
Created attachment 509890 [details]
dmesg when booting with rdblacklist=i915

Booting with rdblacklist=i915. 
vgaswitchero works and let me poweroff the radeon card.

# cat /sys/kernel/debug/vgaswitcheroo/switch
0:DIS: :Off:0000:01:00.0
1:IGD:+:Pwr:0000:00:02.0

I noticed that the screen is unusable after the radeon modeswitching kicks in. The screen is restored to higher resolution later. I guess the i915 driver is doing the modeswitch,  but I'm not sure about it.

Comment 13 Michael Lausch 2011-07-05 07:07:24 UTC
The rdblacklist=radeon kernel command line only works from time to time. Seems to be a timing issue.

Comment 14 Fabian Deutsch 2011-09-22 20:05:23 UTC
I also had  this problem, but this becae very rare with more recent kernels aka. 2.6.40/3.0 under Fedora 15.

Comment 15 Fabian Deutsch 2011-10-27 10:35:07 UTC
Is this still hapening to someone?

Comment 16 David L. Crow 2011-10-27 14:15:06 UTC
I updated to Fedora 16 beta on the machine where I saw this and the problem did not exist there.

Comment 17 Michael Lausch 2011-11-09 14:12:26 UTC
Sorry for responding so late, but i have been away from notebook. Just upgraded to Fedore 16 with all updates and problem still persists.

But i found out one more fact. The T500 has a display port adapter, which is only usable with the radeon card and i have a monitor attached to it. So i can check if the radeon card is active during boot. What i observed multiple times is, that the kernel crashes if the boot screen is displayed on the externel monitor (rendered by the radeon card) but did not crash if the bootscreen is rendered on the notebook screen. Maybe this observation helps.

Comment 18 Michael Lausch 2011-11-23 10:09:22 UTC
Finally i got around to look into this problem again. 
After upgrading to F16 and updating to the newest kernel release, the bug moved into another area. 
It's not udev anymore causing the error, but plymouth. 
I also got the chance to connect a monitor to the display port video output.On the T500 the display port can only be used by the radeon card. What i observed is that the plymouth boot logo appears on the monitor every time the kernel crashes afterwards. radeon.modeset=0 on the kernel command line reliably prevented the kernel to crash. the kernel stack shows that the crash is happening in the close() function of the /dev/dri/card0 file. 

So my conclusion of these facts is:
It depends on the order the cards register themselves. If i915 comes first, /dev/dri/card0 is the i915 card and everything works fine.
If radeon comes first, /dev/dri/card0 is the radeon card and nothing is displayed during boot on the LVD. The muxer connects the LVD to the i915, but plymouth uses the radeon card for rendering and therefore the boot screen is shown on the monitor connected to the display port. 
I don't know if the kernel crash caused by plymouth when closing /dev/dri/card0 is because the driver has a bug, plymouth has a bug, or because the card is disconnected from the LVD.

Comment 19 Fedora End Of Life 2013-02-14 02:30:29 UTC
Fedora 16 changed to end-of-life (EOL) status on 2013-02-12. Fedora 16 is 
no longer maintained, which means that it will not receive any further 
security or bug fix updates. As a result we are closing this bug.

If you can reproduce this bug against a currently maintained version of 
Fedora please feel free to reopen this bug against that version.

Thank you for reporting this bug and we are sorry it could not be fixed.


Note You need to log in before you can comment on or make changes to this bug.