linux-kernel - Re: Apple Thunderbolt Display chaining

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <d40f02e5-7274-e834-6e13-82b3dbe11284@fnarfbargle.com>
Date:   Wed, 30 Mar 2022 22:24:35 +0800
From:   Brad Campbell <lists2009@...rfbargle.com>
To:     Mika Westerberg <mika.westerberg@...ux.intel.com>
Cc:     linux-kernel@...r.kernel.org
Subject: Re: Apple Thunderbolt Display chaining



On 30/3/22 21:43, Mika Westerberg wrote:
> Hi,
> 
> On Wed, Mar 30, 2022 at 09:19:52PM +0800, Brad Campbell wrote:
>> Hey Mika,
>>
>> On 30/3/22 18:18, Mika Westerberg wrote:
>>> Hi,
>>>
>>> On Tue, Mar 29, 2022 at 10:06:35PM +0800, Brad Campbell wrote:
>>>>> Indeed, I did not add this to the "discovery" path yet.
>>>>>
>>>>> I wonder what happens if you change this:
>>>>>
>>>>> +       tunnel = tb_tunnel_alloc_dp(tb, in, out, available_up, available_down,
>>>>> +                                   first ? 0 : 1);
>>>>>
>>>>> to this in your tree:
>>>>>
>>>>> +       tunnel = tb_tunnel_alloc_dp(tb, in, out, available_up, available_down,
>>>>> +                                   first ? 1 : 0);
>>>>>
>>>>
>>>> Here's where it gets all "Apple..y". On the iMac this does the job no matter which
>>>> port the chain is plugged into. Boots and displays correctly first time, every time.
>>>>
>>>> It turns out on the laptop, one port works and the other doesn't. Changing the order
>>>> simply changes which port works. So I assume the EFI sets up the first display using
>>>> the first lane if it's in the first port, and the second if it's in the second.
>>>>
>>>> That means had I managed to perform the first test in the "other port" consistently,
>>>> it would have worked there also.
>>>
>>> Can you try the below patch too? I hard-code the lane based on the
>>> DP adapter number in TBT gen1.
>>>
>>> Let's first figure out proper solution to this issue and then look at
>>> the other one.
>>>
>>> diff --git a/drivers/thunderbolt/tunnel.c b/drivers/thunderbolt/tunnel.c
>>> index a473cc7d9a8d..97d36a7bb527 100644
>>> --- a/drivers/thunderbolt/tunnel.c
>>> +++ b/drivers/thunderbolt/tunnel.c
>>> @@ -865,6 +865,7 @@ struct tb_tunnel *tb_tunnel_alloc_dp(struct tb *tb, struct tb_port *in,
>>
>> This one works from cold boot on both machines regardless of the port the chain is plugged into.
>> It fails on both machines on any hotplug with the symptoms of allocating them both the same link.
>> I added an extra debug into tunnel.c and verified that.
> 
> Hm, okay. What if you change this:
> 
>   link_nr = in->port == 11 ? 1 : 0;
> 
> to this
> 
>   link_nr = in->port == 11 ? 0 : 1;
> 
> Please also keep the debugging you added too so we can see if it now
> uses both lanes.
> 

Nope, that did the same thing. I wonder though. I'm testing it on the laptop and that reports :
[    0.442832] thunderbolt 0000:07:00.0:  Thunderbolt 2 Switch: 8086:156d (Revision: 0, TB Version: 2)

Changing "if (in->sw->generation == 1)" to "if (in->sw->generation == 2)" on the laptop solves that.

I can't test hotplug properly on the iMac due to the radeon training issue.

The laptop still has the issue of a cold boot working in one socket and not the other, but hot plug is working correctly.

Regards,
Brad