Bug 217617

Summary:	[Thunderbolt/USB4, AMD] "path activation failed" on Blackmagic UltraStudio Monitor 3G
Product:	Drivers	Reporter:	Marek Šanta (teslan223)
Component:	Other	Assignee:	drivers_other
Status:	RESOLVED DUPLICATE
Severity:	low	CC:	mario.limonciello, Sanath.S
Priority:	P3
Hardware:	AMD
OS:	Linux
Kernel Version:		Subsystem:
Regression:	No	Bisected commit-id:
Attachments:	dmesg output after connecting the edvice dmesg with thunderbolt.dyndbg='+p'

Description Marek Šanta 2023-06-30 13:37:03 UTC

Created attachment 304510 [details]
dmesg output after connecting the edvice

I am testing the USB4 port on my new AMD laptop. The UltraStudio Monitor 3G device, which works normally on Intel platform (via native Thunderbolt 3) even in older kernel (5.19) has a problem starting up via USB4.

dmesg output is attached - this output just repeats on and on.


Details about my device:
ASUS TUF A15 2023 (FA507NU)
OS: Linux Mint 21.1
Kernel: 6.4 (mainline), with 1 patch applied:
https://lore.kernel.org/linux-acpi/20230601221151.670-1-mario.limonciello@amd.com/raw
CPU: AMD Ryzen 7 7735HS with Radeon Graphics
Cable is standard USB4-certified (it works with Thunderbolt 3 just fine)

Considering that Blackmagic does not support USB4 officially, and also that they apparently do not have the personal capacity to solve Linux problems for 7+ months: ...
https://forum.blackmagicdesign.com/viewtopic.php?f=12&t=173218&p=959939#p959939

... I set the severity to "low" so that only someone interested in running Thunderbolt 3 device on AMD's USB4 would take care. I'm ready to assist with experimenting and applying custom patches.

Comment 1 Sanath S 2023-07-04 04:30:34 UTC

Hi Marek Santa,

Could you please provide a complete log with thunderbolt.dyndbg ?

Comment 2 Marek Šanta 2023-07-04 05:50:37 UTC

Created attachment 304542 [details]
dmesg with thunderbolt.dyndbg='+p'

dmesg output from initialization to a couple of reconnects of the device

Comment 3 Marek Šanta 2023-07-04 05:51:27 UTC

Hi, Sanath, sure, in the attachment - is thunderbolt.dyndbg='+p' what you wanted?

Comment 4 Sanath S 2023-07-13 05:56:19 UTC

Yes.

I dont have the Blackmagic UltraStudio Monitor 3G monitor here with me to debug.
Not able to repro this issue on Samsung/LG monitors I have.

However, from the logs I see some hot unplug events happening as a result lane bonding is getting failed.

Comment 5 Marek Šanta 2023-07-14 12:43:24 UTC

So what would be the next step?
- Are there some modifications in Kernel 6.5 regarding USB4/Thunderbolt? 
- Is it possible to cooperate on a deeper scale? (Would sending a sample of the device to some AMD location help?)
- Or is it more probable that the bug is on the side of Blackmagic?
Thanks

Comment 6 Mario Limonciello (AMD) 2023-08-08 18:01:01 UTC

There are some DMCUB firmware changes that have gone upstream into linux-firmware.git very recently that improve compatibility with TBT3/USB4 displays.

I would suggest upgrading to the most recent from linux-firmware and trying to repro again.

Comment 7 Marek Šanta 2023-08-09 06:41:07 UTC

It is maybe important to say that the device is not a monitor (a display) - it is a generic PCIe device for outputting 3G-SDI signal:
https://www.blackmagicdesign.com/products/ultrastudio/techspecs/W-DLUS-13

Related issue is occurring with Intel 12-gen now
(but it works well on 10-gen and 11-gen):
https://bugzilla.kernel.org/show_bug.cgi?id=217737

But yes, I have tried latest linux-firmware.git (0ab353f8) now but nothing changes :/

Comment 8 Mario Limonciello (AMD) 2023-08-09 11:37:14 UTC

Got it; in that case I feel these are duplicate issues.  The newer Intel gens use the Linux USB4 SW connection manager provided by thunderbolt.ko just like the AMD does.  I'll mark a duplicate accordingly as there is a lot more debug going on that issue.

*** This bug has been marked as a duplicate of bug 217737 ***