Message-ID: <7c9474d2-d948-4d1d-6f7b-94335b8b1f15@fnarfbargle.com>
Date:   Wed, 28 Aug 2019 17:12:00 +0800
From:   Brad Campbell <lists2009@...rfbargle.com>
To:     Mika Westerberg <mika.westerberg@...ux.intel.com>
Cc:     linux-kernel@...r.kernel.org, michael.jamet@...el.com,
        YehezkelShB@...il.com
Subject: Re: Thunderbolt DP oddity on v5.2.9 on iMac 12,2

On 28/8/19 3:33 pm, Mika Westerberg wrote:
> On Wed, Aug 28, 2019 at 03:09:07PM +0800, Brad Campbell wrote:
>> G'day All,
> 
> Hi,
> 
>> 5.2 is the first kernel that ha/thunderbolt

s allowed me to use 2 Apple Thunderbolt
>> display on a 2011 vintage iMac. Awesome work and many thanks!
>>
>> I boot this machine in BIOS (Bootcamp) mode as that gave me brightness
>> control over the internal panel.
>>
>> I've been using it with a Thunderbolt display and a standard DVI display
>> since 2012 as the Apple firmware would only set up a single Thunderbolt
>> display at bootup. Didn't find that out until I'd bought a second one and
>> kept it around as a spare.
>>
>> When I saw the changelog for 5.2 I thought I'd see if it would run 2
>> Thnunderbolt displays, and it does after a minor fiddle. I have tried this
>> with the Kanex adapter unplugged just for completeness. No change.
>>
>> When I boot the machine it picks up both displays :
>>
>> [    1.072080] thunderbolt 0-1: new device found, vendor=0x1 device=0x8002
>> [    1.072086] thunderbolt 0-1: Apple, Inc. Thunderbolt Display
>> [    1.392076] random: crng init done
>> [    1.595609] thunderbolt 0-3: new device found, vendor=0x1 device=0x8002
>> [    1.595615] thunderbolt 0-3: Apple, Inc. Thunderbolt Display
>> [    1.905117] thunderbolt 0-303: new device found, vendor=0xa7 device=0x4
>> [    1.905121] thunderbolt 0-303: Kanex Thunderbolt to eSATA+USB3.0 Adapter
>> [    1.910098] thunderbolt 0000:07:00.0: 0:c: path does not end on a DP
>> adapter, cleaning up
>>
>> Display 0-3 works. Display 0-1 remains blank.
>>
>> If I unplug 0-1 and re-plug it, it is detected and comes up perfectly.
> 
> Can you add "thunderbolt.dyndbg" to the kernel command line and
> reproduce this? That gives bit more information what might be happening.

dmesg.03 attached.

> I'm suspecting that the boot firmware does configure second DP path also
> and we either fail to discover it properly or the boot firmware fails to
> set it up.
> 
> Also if you boot with one monitor connected and then connect another
> (when the system is up) does it work then?

Umm.. so this is where it gets weird. No it doesn't. Apparently it fails 
to configure the first monitor it finds. This is the one the Apple 
bootcamp firmware configures at boot.

So if I disconnect 0-1 and boot up, it detects 0-3 but it doesn't work 
(same way 0-1 doesn't work). A plug-unplug brings that up and then 
plugging in 0-1 brings that up too.

I then did a few swapsies trying to get the heads in the right order and 
gave up and rebooted.

This is in dmesg.4 attached.

dmesg.3 is the straight boot with dyndbg enabled and both heads plugged 
in. dmesg.4 is with 0-3 plugged in only. It comes up dead and then I 
play swapsies getting things working.

Took me a while to figure out I didn't have CONFIG_DYNAMIC_DEBUG in the 
kernel.

Now it makes sense. I used to have a Thunderbolt display and a DVI 
monitor with an adapter. < v5.2 both of those worked just fine. When I 
tried 5.2 I had to unplug/replug the Thunderbolt display to make it 
work. When I added the second Thunderbolt display, it took the place of 
the DVI display and is the first one the firmware tries to configure.

I wonder if I boot it up using EFI the firmware won't try and configure 
either of them, but then I'll lose my backlight control. I'll give that 
a go anyway.

Regards,
Brad

View attachment "dmesg.03" of type "text/plain" (130299 bytes)

View attachment "dmesg.04" of type "text/plain" (143897 bytes)

Powered by blists - more mailing lists