Bug 45421

Summary: regression: nouveau.ko fails to load
Product: xorg Reporter: Jesse Brandeburg <jesse.brandeburg>
Component: Driver/nouveauAssignee: Nouveau Project <nouveau>
Status: CLOSED FIXED QA Contact: Xorg Project Team <xorg-team>
Severity: major    
Priority: medium    
Version: git   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
dmesg from boot for failed driver load
none
dmesg with drm.debug=14 from boot with 3.3.0-rc1
none
vbios from HP 8530w/770M
none
8530w acpi dump (tarbz2) as requested
none
acpi dump after cp to new directory
none
output from driver after patch to make it load none

Description Jesse Brandeburg 2012-01-30 18:23:03 UTC
Created attachment 56355 [details]
dmesg from boot for failed driver load

nouveau fails to load
v3.2 works okay
v3.3-rc1 fails
nouveau git fails

I tried to bisect but got compile errors due to HWMON, and then two panics in a row on different commits, so I gave up bisect.

full dmesg from boot attached, the critical bit is:
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.455869] MXM: GUID detected in BIOS
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.467051] VGA switcheroo: detected Optimus DSM method \ handle
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.481163] [drm] nouveau 0000:01:00.0: Detected an NV50 generation card (0x096e00a1)
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.492671] checking generic (d1000000 1e0000) vs hw (c0000000 10000000)
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.492673] checking generic (d1000000 1e0000) vs hw (d0000000 2000000)
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.492675] fb: conflicting fb hw usage nouveaufb vs VESA VGA - removing generic driver
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.504429] Console: switching to colour dummy device 80x25
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.510526] sdhci-pci 0000:86:09.1: Will use DMA mode even though HW doesn't fully claim to support it.
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.513433] [drm] nouveau 0000:01:00.0: Attempting to load BIOS image from PRAMIN
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.579993] sdhci-pci 0000:86:09.1: Will use DMA mode even though HW doesn't fully claim to support it.
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.599872] [drm] nouveau 0000:01:00.0: ... appears to be valid
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.599876] [drm] nouveau 0000:01:00.0: BIT BIOS found
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.599880] [drm] nouveau 0000:01:00.0: Bios version 62.94.3b.00
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.599883] [drm] nouveau 0000:01:00.0: TMDS table version 2.0
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.600121] [drm] nouveau 0000:01:00.0: MXM: BIOS version 2.1
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.600449] [drm] nouveau 0000:01:00.0: MXM: failed to locate valid SIS
Jan 30 17:07:13 jbrandeb-mobl kernel: [    2.605529] nouveau: probe of 0000:01:00.0 failed with error -22
Comment 1 Ben Skeggs 2012-01-30 19:16:52 UTC
Thanks for the report.

Firstly, are you able to boot with "drm.debug=14 log_buf_len=1M" for me, and attach the resulting dmessage.

Secondly, can you also attach /sys/kernel/debug/dri/0/vbios.rom for me please (load "working" nouveau, and mount debugfs at /sys/kernel/debug to get it).

And finally, could you also copy all the files from /sys/firmware/acpi/tables somewhere, and tar them up and attach them here.

Thanks!
Ben.
Comment 2 Jesse Brandeburg 2012-01-31 09:23:22 UTC
Created attachment 56400 [details]
dmesg with drm.debug=14 from boot with 3.3.0-rc1
Comment 3 Jesse Brandeburg 2012-01-31 09:30:40 UTC
Created attachment 56401 [details]
vbios from HP 8530w/770M
Comment 4 Jesse Brandeburg 2012-01-31 09:31:25 UTC
Created attachment 56402 [details]
8530w acpi dump (tarbz2) as requested
Comment 5 Jesse Brandeburg 2012-01-31 17:10:56 UTC
Created attachment 56420 [details]
acpi dump after cp to new directory

sorry if the files were zero before, didn't check that.  Files have content now.
Comment 6 Ben Skeggs 2012-01-31 21:35:00 UTC
(In reply to comment #5)
> Created attachment 56420 [details]
> acpi dump after cp to new directory
> 
> sorry if the files were zero before, didn't check that.  Files have content
> now.

Thanks!  I believe I know where the root cause of this issue is, and I'll attempt to get a patch ready to fix it properly tomorrow.

In the meantime, nouveau git will no longer completely fail init in these situations, so you should be able to use it again.  The patch has been sent upstream for 3.3-rc also.
Comment 7 Jesse Brandeburg 2012-02-01 15:43:23 UTC
Created attachment 56482 [details]
output from driver after patch to make it load
Comment 8 Jesse Brandeburg 2012-02-08 09:41:08 UTC
nouveau git commit that seems to fix it: f41a8e1bcbd69148dd39568a6f062ae698ad4afc

upstream commit that works around the issue (I think): ce2e7895faba8fabaa917f52293126e5f4174fa9

thanks Ben!!!

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.