Problems with kernel-3.8 and kernel-3.9 modesetting

[rockdoctor]

I don't know if the problem lies with the kernel's nouveau module, Xorg's nouveau driver, or both of my older onboard Nvidia graphics systems (9100 and 6150LE), but any time I try to boot a version 3.8 or 3.9 kernel, once modesetting kicks in, the signal to the monitor goes away and there's no switching to a tty. The vesa driver works, but won't cut it with my 1680x1050 monitor. Kernel versions 3.7 and earlier work fine, whereas the 3.8 and 3.9 kernels fail to properly modeset (is that a real verb?) when used on F18 and F17 systems. Anyone else seeing this?

[matthewjavelet]

i'm experiencing the same issues. i'm not using anything nvidia though, intel hd graphics here.

[Colors]

The same issue on a Compaq CQ58 with intel hd and kernel 3.7.x. I use vesa and nomodeset in grub as workaround. With kernel 2.6.10-4 no problem.

[rockdoctor]

Looking at /var/log/messages, the third line in the messages generated by kernel-3.9 appears to be the error message associated with the condition.

3.9 kernel (fails):
Feb 26 08:25:42 Pyrope kernel: [ 7.153565] nouveau 0000:02:00.0: VGA-1: EDID block 0 invalid.
Feb 26 08:25:42 Pyrope kernel: [ 7.153583] nouveau E[ DRM] DDC responded, but no EDID for VGA-1
Feb 26 08:25:42 Pyrope kernel: [ 7.154651] nouveau 0000:02:00.0: No connectors reported connected with modes
Feb 26 08:25:42 Pyrope kernel: [ 7.154718] [drm] Cannot find any crtc or sizes - going 1024x768
Feb 26 08:25:42 Pyrope kernel: [ 7.156635] nouveau [ DRM] allocated 1024x768 fb: 0x50000, bo ffff8801108a7430
Feb 26 08:25:42 Pyrope kernel: [ 7.159120] fbcon: nouveaufb (fb0) is primary device
Feb 26 08:25:42 Pyrope kernel: [ 7.159987] Console: switching to colour frame buffer device 128x48
Feb 26 08:25:42 Pyrope kernel: [ 7.160471] nouveau 0000:02:00.0: fb0: nouveaufb frame buffer device
Feb 26 08:25:42 Pyrope kernel: [ 7.160482] nouveau 0000:02:00.0: registered panic notifier
Feb 26 08:25:42 Pyrope kernel: [ 7.160531] [drm] Initialized nouveau 1.1.0 20120801 for 0000:02:00.0 on minor 0

3.7 kernel (works properly):
Feb 26 08:30:01 Pyrope kernel: [ 4.673034] nouveau 0000:02:00.0: VGA-1: EDID block 0 invalid.
Feb 26 08:30:01 Pyrope kernel: [ 4.673039] nouveau E[ DRM] DDC responded, but no EDID for VGA-1
Feb 26 08:30:01 Pyrope kernel: [ 4.719650] nouveau [ DRM] allocated 1024x768 fb: 0x50000, bo ffff880114c60000
Feb 26 08:30:01 Pyrope kernel: [ 4.719730] fbcon: nouveaufb (fb0) is primary device
Feb 26 08:30:01 Pyrope kernel: [ 4.719801] nouveau E[ DRM] no native mode, forcing panel scaling
Feb 26 08:30:01 Pyrope kernel: [ 4.737449] Console: switching to colour frame buffer device 128x48
Feb 26 08:30:01 Pyrope kernel: [ 4.737839] fb0: nouveaufb frame buffer device
Feb 26 08:30:01 Pyrope kernel: [ 4.737854] drm: registered panic notifier
Feb 26 08:30:01 Pyrope kernel: [ 4.737871] [drm] Initialized nouveau 1.1.0 20120801 for 0000:02:00.0 on minor 0

[AdamW]

Messages like "The same issue on a Compaq CQ58 with intel hd and kernel 3.7.x" always tickle me

So...it's the same issue but on a different kernel (which the OP says works for him), with a different graphics card from a different manufacturer?

That is not the same issue. =)

rockdoctor: I'm using a 3.9 kernel on Rawhide with a Geforce 9600 GT, not seeing the problem. So it's hardware-specific to some degree. Can you file a bug for this, against xorg-x11-drv-nouveau , including the logs described at https://fedoraproject.org/wiki/How_t..._Xorg_problems ? Thanks!

[rockdoctor]

Already filed - https://bugzilla.redhat.com/show_bug.cgi?id=910945 Problem is that I can't see anything once the screen goes black. Added smolt profile from working boot.

[AdamW]

rockdoctor: funnily enough, I'm now seeing a very similar issue with 3.9.rc0.git11 - but it's not the same as yours, because 3.9.rc0.git7 works fine. Hardware, eh. Sigh.

[hadrons123]

try acpi_backlight=vendor in your kernel command line

[rockdoctor]

Well, I moved up to kernel-3.9.0-0.rc0.git8.1.fc19.x86_64, and while it's still problematic, instead of a black screen I get in text mode) bunches of white pixels that look like they might represent a mismapping of the pixels would ordinarily form legible text.

---------- Post added at 08:58 AM ---------- Previous post was at 07:55 AM ----------

Adam: Interestingly, kernel-3.9.0-0.rc0.git11 behaves the same as .3.9.0-0.rc0.git8 on my hardware (I never tried the git7 kernel). The scrambled pixels are definitely a misrepresentation of the expected text; I can see the bright green of my prompt. I captured /var/log/Xorg.0.log and the relevant portion of /var/log/messages by booting off a (full) flash drive install of F17. If you'll file the bug (so it starts out with all the proper info), I'll contribute my data

Hadrons: acpi_backlight=vendor didn't help (tried it on an Ubuntu 13.04 live USB as kernel-3.8 was purged from my Rawhide install yesterday

[AdamW]

rockdoctor: our two problems are clearly not the same, so we should file different bugs. (airlied says the big DRM merge for 3.9 hit at rc0.git8, so *my* problem is likely caused by the DRM merge. It sounds like your problem came in between 3.7 and 3.8, and the 3.9 DRM merge actually improves things somewhat for you, but still doesn't make it right.)

[rockdoctor]

https://bugzilla.redhat.com/show_bug.cgi?id=917260 Hope I attached all the relevant files

[VernDog]

This appears to be the same problem I am having, and not only with Fedora, but with any KDE system. Fedora was the last one I installed, so I tried using modesetting=0 and it now works. I had no clue what was causing the freezing.

I found the only way to stop the freezing, was just before the final "big K" icon (I don't know what those 5 icons that dance across the KDE screen - tool wrench, world, then final the Big K). At any rate just beofre the Big-K I would start a virtual terminal , then it would boot up normally. If I didn't do that , then the screen would loose horizonal and everything would freeze. On occasion it would boot up normally.

The xsession log would increase to over 20megs and counting. Over 50,000 lines.

I have a AMD 64bit, dual core with NVIDIA GeForce 6150SE nForce 430 integrated video.

It never occured to me that it was the kernel causing the problem. I thought it was my nvidia video. I missed the last release of KDE. Ubuntu 12.04 and even the current 13.10 all work using Unity.

I guess I should head over to the bug report and relay my problem as well, since its related.

Thanks for this topic!