Linux status with Thunderbolt

Talk about your MIDI interfaces, microphones, keyboards...

Moderators: khz, MattKingUSA

Post Reply
retnev
Established Member
Posts: 95
Joined: Sat Mar 25, 2017 2:13 am

Linux status with Thunderbolt

Post by retnev »

Is there any success with developing a thunderbolt driver for Linux. .
Any new updates ?

Last I saw was that the 5.2 Kernel should have support. The 5.4 Kernel was released in Nov 2019.
Anyone tried it ?
Adding to the excitement of the Linux 5.2 kernel changes are a lot of Thunderbolt improvements expected to be introduced in this next kernel cycle.
Mika Westerberg of Intel has been working on a lot of Thunderbolt connectivity improvements destined for Linux 5.2 and in recent days has begun staging this work in the thunderbolt-next tree ahead of the Linux 5.2 kernel merge window opening in May.
This work includes support for full PCIe daisy chaining within the Thunderbolt software connection manager:
Currently the software connection manager (tb.c) has only supported creating a single PCIe tunnel, no PCIe device daisy chaining has been supported so far. This updates the software connection manager so that it now can create PCIe tunnels for full chain of six devices.

User avatar
Be.
Established Member
Posts: 41
Joined: Mon Mar 16, 2015 4:51 am

Re: Linux status with Thunderbolt

Post by Be. »

Linux supports Thunderbolt, but AFAIK no one has written a Linux driver for a Thunderbolt audio interface yet. Unlike USB, there is no standard for Thunderbolt audio devices, so each device would need its own driver.

retnev
Established Member
Posts: 95
Joined: Sat Mar 25, 2017 2:13 am

Re: Linux status with Thunderbolt

Post by retnev »

Oh lordy, no standard !
As I can remember it was actually apple that forced the USB compliance Standard.
But, now since this is their invention and USB wasnt their invention, dont bet on it that they will demand a standard this time.

Jamesf
Established Member
Posts: 44
Joined: Tue May 28, 2019 12:23 pm
Location: Madrid

Re: Linux status with Thunderbolt

Post by Jamesf »

retnev wrote:
Sun Jun 28, 2020 3:34 am
Oh lordy, no standard !
As I can remember it was actually apple that forced the USB compliance Standard.
But, now since this is their invention and USB wasnt their invention, dont bet on it that they will demand a standard this time.
Thunderbolt was actually an Intel creation.
Apple adopted it early, which is why their name was associated with it. So they might, but don't hold your breath.

I'd like to think the industry would have learned long ago that trying to corner the market with your proprietary solution is almost always a losing game (anybody remember DECnet, pre-ATAPI HDDs, or Compaq?), but every generation seems convinced it'll be different this time.
I'm slow, but I get there eventually.

retnev
Established Member
Posts: 95
Joined: Sat Mar 25, 2017 2:13 am

Re: Linux status with Thunderbolt

Post by retnev »

Sorry about that, you are right, I stand corrected.

It is dumbfounding that every new interface protocol go through the same failed bump of the head expecting a different outcome.
History is lost on them it seems.

You left out microsoft (proprietary) who now have to scramble against Linux (open). Their stubborn clinging to things like the proprietary WINSOCK lol ! while Linux and Unix already had 65536 filtered ports. That was a proprietary hole they reamed for themselves and had to lift the entire open source kerberos and filter system from BSD Unix under BSD license.
Look what happened to the proprietary Un*xes. Only BSD (which is open) basically still standing.

Yet every time they try do it the same again.

I will see if I can get a thunderbolt interface and attempt a driver.

User avatar
wrl
Established Member
Posts: 40
Joined: Wed Nov 03, 2010 12:46 am

Re: Linux status with Thunderbolt

Post by wrl »

I'd be happy to share notes. I have an 828es connected via Thunderbolt, passed-through to a qemu win10 VM using vfio. I'm able to capture PCIe reads/writes, but I've not yet been able to capture the DMA mappings for the actual audio data. mmiotrace is similarly unable to capture it. From a read of the qemu source, there appears to be a mapping operation that mmaps system RAM through into the guest, and then qemu is no longer involved in the i/o. I reckon the device is getting some sort of interrupt when buffers are ready, but it could be a big ringbuffer or similar. No idea yet.

Nonetheless, the setup is showing some promise. I'm a PCIe neophyte so I've been having to get myself up to speed on the bus in general.

Jamesf
Established Member
Posts: 44
Joined: Tue May 28, 2019 12:23 pm
Location: Madrid

Re: Linux status with Thunderbolt

Post by Jamesf »

Sorry about my tone before; I didn't realise how curt that comment sounded, until I re-read it just now.
retnev wrote:
Mon Jun 29, 2020 5:32 am
It is dumbfounding that every new interface protocol go through the same failed bump of the head expecting a different outcome.
History is lost on them it seems.
Sure does look that way.

retnev wrote:
Mon Jun 29, 2020 5:32 am
You left out microsoft...
Mostly because, when you've been around this long, Microsoft is a whole other category of rants and it's kinda hard to rein it in :)
I'll give them credit, though: Satya Nadella has really turned that company around, and given it a whole new attitude. I'm at the point now where I'd pick Windows over MacOS if I had to choose one of those for audio work.
I've gotten past my disorientation at realising that Linux has become the OS that "just works", Microsoft became the UI pioneers, and MacOS is the one that you don't upgrade until the third patch-release.
retnev wrote:
Mon Jun 29, 2020 5:32 am
I will see if I can get a thunderbolt interface and attempt a driver.
I'll be watching with interest.
I'm slow, but I get there eventually.

retnev
Established Member
Posts: 95
Joined: Sat Mar 25, 2017 2:13 am

Re: Linux status with Thunderbolt

Post by retnev »

Been trying to get an affordable interface to try a driver, but they all end up expensive. Will have to wait.

Post Reply