Bug 104971

Summary:	Regression on thunderbolt display on Macs
Product:	Drivers	Reporter:	Radu Andries (admiral0)
Component:	Video(DRI - Intel)	Assignee:	Greg Kroah-Hartman (greg)
Status:	RESOLVED CODE_FIX
Severity:	normal	CC:	intel-gfx-bugs, kernel-bugs
Priority:	P3
Hardware:	All
OS:	Linux
Kernel Version:	4.1.6	Subsystem:
Regression:	No	Bisected commit-id:
Attachments:	Dmesg and lspci Dmesg with 3.18.20 3.18.20 - good with drm.debug=14 4.1.6 - doesn't work with drm.debug=14

Description Radu Andries 2015-09-25 09:20:04 UTC

Created attachment 188351 [details]
Dmesg and lspci

A display connected via a thunderbolt was working with kernel 3.17 but is not detected with kernel 4.1

The issue is also documented here -> https://bugzilla.redhat.com/show_bug.cgi?id=1260796

Attaching logs (dmesg and lspci in the same file)

Comment 1 Radu Andries 2015-09-25 12:27:25 UTC

Created attachment 188411 [details]
Dmesg with 3.18.20

Good dmesg (Second monitor works)

Comment 2 Greg Kroah-Hartman 2015-09-25 17:15:59 UTC

Please send to the graphics/video driver people, this is not a "hotplug"
issue.

Comment 3 Radu Andries 2015-09-28 07:12:23 UTC

Well, I think it's the thunderbolt module. There is no such component in bugzilla.

Comment 4 Jani Nikula 2015-09-28 08:26:05 UTC

Nothing obvious that's wrong in dmesg, please add drm.debug=14 module parameter and attach the dmesg. A bisect result would be great.

Comment 5 Radu Andries 2015-09-29 08:38:09 UTC

Created attachment 188931 [details]
3.18.20 - good with drm.debug=14

Comment 6 Radu Andries 2015-09-29 08:39:10 UTC

Created attachment 188941 [details]
4.1.6 - doesn't work with drm.debug=14

Comment 7 Radu Andries 2015-09-29 08:40:28 UTC

I can bisect this weekend (and hack a rough patch probably) if this is not enough.

Comment 8 Jani Nikula 2015-10-07 12:19:45 UTC

Please try v4.3-rc, I think this may be fixed upstream.

Comment 9 Radu Andries 2015-10-08 07:21:57 UTC

Yes, I confirm 4.3-rc3 fixes this issue.

Comment 10 Jani Nikula 2015-10-08 11:10:08 UTC

(In reply to Radu Andries from comment #9)
> Yes, I confirm 4.3-rc3 fixes this issue.

Thanks for testing, closing.

There are *many* commits which may have alone, or in some combination, fixed the issue. Some of them may not even be suitable for backporting to stable kernels. Since it's fixed upstream, I'm afraid we're not putting in the effort to identify the commits that fixed the issue.

However, if it's of value to you to have the fixes backported, you should do a reverse git bisect and find out, and then let us know. Alternatively try backporting the commits referencing i2c or aux in drivers/gpu/drm/drm_dp_helpers.c and drivers/gpu/drm/i915/intel_dp.c, I'm fairly sure that's where the fix is.

Thanks.