Intel Thunderbolt Is Still A Pain For Linux Developers, Users
While Thunderbolt was promising from a technology perspective, it hasn't seen too much adoption outside of Apple systems and the Linux support is still plaguing developers and causing nightmares among Linux users.
In the past on Phoronix I have written many times about Thunderbolt support. Two years ago I wrote about the problems in using an Apple Thunderbolt Display on Linux and since then it's only recently started getting better. From a user perspective, Thunderbolt is wonderful: a single cable being able to transfer a high resolution display, audio, USB connectivity, and Ethernet network support. Up until recently when I stopped using my Retina MacBook Pro, it was wonderfully convenient connecting it to the Apple Thunderbolt Display with just having to connect a Thunderbolt cable and power cable. There's also other interesting possibilities with daisy chaining displays and other possibilities for this lightning-fast interface using PCI-E.
Last year in the Linux kernel there was some work on Apple Thunderbolt hot-plugging support but that was the last major update I've had to share with Phoronix readers. Fortunately, at the Linux Foundation Collaboration Summit in Napa Valley, Rafael J. Wysocki of Intel and the Linux kernel's ACPI / PM subsystem maintainer, is speaking on ACPI-based support for Thunderbolt hot-plugging.
I'll be at the session later in person, but there's already out some slides from this presentation. Those wishing to see a sneak preview of what's going to be said, you can find the PDF slides via this link. For ACPI-based Thunderbolt hot-plug support, Wysocki says it should work on systems where the BIOS participates in the handling of link events. The Linux kernel code for handling the ACPI bus check notifications was "modified substantially" for supporting Thunderbolt hot-plugging, the PCI subsystem locking was hardened in the process, and the Thunderbolt hot-plug support for it to work on Linux has to assume the device's driver is prepared to handle the device being removed from the system.
In the past on Phoronix I have written many times about Thunderbolt support. Two years ago I wrote about the problems in using an Apple Thunderbolt Display on Linux and since then it's only recently started getting better. From a user perspective, Thunderbolt is wonderful: a single cable being able to transfer a high resolution display, audio, USB connectivity, and Ethernet network support. Up until recently when I stopped using my Retina MacBook Pro, it was wonderfully convenient connecting it to the Apple Thunderbolt Display with just having to connect a Thunderbolt cable and power cable. There's also other interesting possibilities with daisy chaining displays and other possibilities for this lightning-fast interface using PCI-E.
Last year in the Linux kernel there was some work on Apple Thunderbolt hot-plugging support but that was the last major update I've had to share with Phoronix readers. Fortunately, at the Linux Foundation Collaboration Summit in Napa Valley, Rafael J. Wysocki of Intel and the Linux kernel's ACPI / PM subsystem maintainer, is speaking on ACPI-based support for Thunderbolt hot-plugging.
I'll be at the session later in person, but there's already out some slides from this presentation. Those wishing to see a sneak preview of what's going to be said, you can find the PDF slides via this link. For ACPI-based Thunderbolt hot-plug support, Wysocki says it should work on systems where the BIOS participates in the handling of link events. The Linux kernel code for handling the ACPI bus check notifications was "modified substantially" for supporting Thunderbolt hot-plugging, the PCI subsystem locking was hardened in the process, and the Thunderbolt hot-plug support for it to work on Linux has to assume the device's driver is prepared to handle the device being removed from the system.
18 Comments