Bug 73791

Summary: [NV96] [BISECTED] Regression: 9600M GT fails to boot with 3.13-rc8
Product: xorg Reporter: Andreas Reis <andreas.reis>
Component: Driver/nouveauAssignee: Nouveau Project <nouveau>
Status: RESOLVED FIXED QA Contact: Xorg Project Team <xorg-team>
Severity: critical    
Priority: medium    
Version: git   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
dmesg (rc8 with commit reverted)
none
dmesg of bugged 3.13-rc8+ kernel
none
dmesg of kernel 3.12.7
none
pass mxm pointer separately none

Description Andreas Reis 2014-01-19 12:35:53 UTC
Created attachment 92385 [details]
dmesg (rc8 with commit reverted)

Commit "populate master subdev pointer only when fully constructed" breaks support for my 9600M GT (G96 on an Acer Aspire 6935G, bios 1.21), identified by bisection and reverting it from the current git head (7d0d46da).

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=61b365a505d625734be9ac54066fe4700672abc9

Instead of the framebuffer switching to the display's native resolution it just stays at the one given by the bios and the boot comes to a halt after a few seconds (with a few tasks, like wifi initialization, still being done afterwards).

(I don't know the proper terminology; I hope that this description is clear enough…)
Comment 1 Ilia Mirkin 2014-01-19 12:42:57 UTC
Did the kernel that you were testing contain https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=fdd239ac99a0cc298b382c5ab5e7bcd09e8933d7 ?
Comment 2 Andreas Reis 2014-01-19 12:48:53 UTC
It's at Torvalds's current HEAD ("Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net"), so yes.

I also double checked 'git diff'; the 'populate master' commit is the only one reverted.
Comment 3 Ilia Mirkin 2014-01-19 12:57:47 UTC
The only weird thing I see in your log is

[    0.766097] nouveau  [     MXM][0000:01:00.0] unmatched output device 0x96006403e8060100
[    0.766099] nouveau  [     MXM][0000:01:00.0] unmatched output device 0x0000004100009600

Can you get the dmesg from one of the failed boots with a HEAD kernel? It's probably in your system logs. That should pinpoint exactly where the issue is.
Comment 4 Andreas Reis 2014-01-19 14:26:32 UTC
Created attachment 92395 [details]
dmesg of bugged 3.13-rc8+ kernel

Sorry for the delay; took me a while to remember adding "emergency" to the kernel line in order to be able to copy the faulty dmesg.

I recompiled the kernel (at said "Merge ... net", ie. with the commit you mentioned included). Turns out there's still a NULL pointer deference somewhere.
Comment 5 Andreas Reis 2014-01-19 14:37:29 UTC
Created attachment 92398 [details]
dmesg of kernel 3.12.7

There's the dmesg of the (trouble-free) kernel 3.12.7 for good measure; the two "unmatched output devices" are already there.

Speaking of which, maybe that's related to the HDMI issue: I can connect my old ViewSonic VX924 via VGA just fine, but when I try to connect it via HDMI (via DVI converter cable) it gets recognized as DVI-D-1 just fine (ie. with proper resolutions, etc.), but doesn't receive any input upon activation. No such issue with the proprietary nvidia driver, but that's obviously for another bug report.
Comment 6 Ilia Mirkin 2014-01-19 14:48:10 UTC
Created attachment 92399 [details] [review]
pass mxm pointer separately

For your HDMI issue -- can you see if the patch at https://bugs.freedesktop.org/attachment.cgi?id=92185 helps (and if so, mention something to that effect in bug #60680).

For your MXM issue -- does this help?
Comment 7 Andreas Reis 2014-01-19 15:18:57 UTC
Both fixes work properly, thanks!

(The two "unmatched output device" lines are the same as before, though.)
Comment 8 Ilia Mirkin 2014-01-20 03:34:11 UTC
Looks like my patch made it into 3.13 (commit 72de182362... last one!). Thanks for the report.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.