Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:18 as permitted sender) client-ip=2620:137:e000::1:18;
Date:   Wed, 6 Apr 2022 17:56:50 +0300
From:   Mika Westerberg <mika.westerberg@linux.intel.com>
To:     Brad Campbell <lists2009@fnarfbargle.com>
Cc:     linux-kernel@vger.kernel.org
Subject: Re: Apple Thunderbolt Display chaining
Message-ID: <Yk2qMt568oEeTj8H@lahna>
References: <YkRtf2HiXBOWGjna@lahna>
 <eacc22dc-e650-a208-9484-62edd98d5778@fnarfbargle.com>
 <YkXYdcUZ1E18ENle@lahna>
 <5dcee6f7-cc8c-e3ce-920c-4ad3f5d77e14@fnarfbargle.com>
 <YkcMofXKr/ZmOSwH@lahna>
 <7ea44c20-6c65-224f-af7b-aa1bd310d038@fnarfbargle.com>
 <YkrEE0uh9EVCchfl@lahna>
 <d112651d-919d-ce2d-efac-6abbd9de0f07@fnarfbargle.com>
 <YkrqL/wnACNVOpSi@lahna>
 <ae336e2d-8e7c-9167-ab3d-8d642cd4fb2c@fnarfbargle.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <ae336e2d-8e7c-9167-ab3d-8d642cd4fb2c@fnarfbargle.com>
Organization: Intel Finland Oy - BIC 0357606-4 - Westendinkatu 7, 02160 Espoo
Precedence: bulk

Hi,

On Wed, Apr 06, 2022 at 10:51:41AM +0800, Brad Campbell wrote:
> Both included in-line.
> 
> This is cold boot with the chain plugged in. I've re-added the dbg to
> print the link number, and I've included your path discovery debugs.
> Boot with chain plugged in, wait for it to settle, unplug and replug.
> First head in the chain fails with :
> 
> [   65.778129] [drm:radeon_dp_link_train [radeon]] *ERROR* displayport link status failed
> [   65.778158] [drm:radeon_dp_link_train [radeon]] *ERROR* clock recovery failed

Thanks for the logs! 

The DP tunnels look pretty much the same except that the Apple EFI CM
seems to assign 7 buffers for the AUX RX path first hop whereas we
always use 1 buffer. Not sure if that really makes a difference and we
could try to use the same number but first, I realized that the PCI
resource allocation seems not to work properly.

Can you disable PCIe tunneling (if you use Ubuntu/Fedora or similar
there is the "Thunderbolt -> Direct Access" switch that you can turn
off) and try again? Please also take 'sudo lspci -vv' for the resulting
topology. I suspect this might also affect the other issues (the
timeouts) you are seeing. Note this makes the peripherals connected to
the monitors unusable too.