Bugzilla – Bug 1212587
Kernel 6.3.7-1.2 fails to activate Intel Thunderbolt PCIe Bridge
Last modified: 2023-07-12 11:44:25 UTC
Created attachment 867722 [details] Boot log on affected kernel After updating to the latest Tumbleweed kernel (v6.3.7-1.2), the USB devices connected through my Dell laptop USB-C docking station stopped working. If I reboot to an older kernel (6.2.10-1.1), they work fine. Comparing lspci output from the two kernels shows that the newer kernel fails to activate the following devices: 03:00.0 PCI bridge: Intel Corporation JHL6340 Thunderbolt 3 Bridge (C step) [Alpine Ridge 2C 2016] (rev 02) (prog-if 00 [Normal decode]) Subsystem: Device 2222:1111 04:00.0 PCI bridge: Intel Corporation JHL6340 Thunderbolt 3 Bridge (C step) [Alpine Ridge 2C 2016] (rev 02) (prog-if 00 [Normal decode]) Subsystem: Device 2222:1111 04:01.0 PCI bridge: Intel Corporation JHL6340 Thunderbolt 3 Bridge (C step) [Alpine Ridge 2C 2016] (rev 02) (prog-if 00 [Normal decode]) Subsystem: Device 2222:1111 04:02.0 PCI bridge: Intel Corporation JHL6340 Thunderbolt 3 Bridge (C step) [Alpine Ridge 2C 2016] (rev 02) (prog-if 00 [Normal decode]) Subsystem: Device 2222:1111 3b:00.0 USB controller: Intel Corporation JHL6340 Thunderbolt 3 USB 3.1 Controller (C step) [Alpine Ridge 2C 2016] (rev 02) (prog-if 30 [XHCI]) Subsystem: Device 2222:1111
Created attachment 867723 [details] lspci -vv from kernel 6.2.10-1.1
Created attachment 867724 [details] lspci -vv from kernel 6.3.7-1.2
Could you check with the latest package in OBS Kernel:stable repo? A TB fix patch from 6.4-rc7 is included there. Also, please check whether the early 6.3.x kernel also suffers. At least 6.3.2 is still found in TW history repo.
Created attachment 867725 [details] lspci -vv from kernel 6.3.2-1.1 Kernel v6.3.2-1.1: Not affected
Created attachment 867726 [details] lspci -vv from kernel 6.3.8-4.1.ge044c8e Kernel v6.3.8-4.1.ge044c8e (Kernel:stable): Affected
Thanks. At least it's a good news that 6.3.2 worked. There are 6.3.4 and 6.3.6 kernels in the history repo. Could you verify which one starts breaking (or not), too?
Created attachment 867727 [details] lspci -vv from kernel 6.3.4-1.1 Both kernels v6.3.4-1.1 and v6.3.6-1.1 are affected.
Thanks. So the likely cause is the fix patch for AMD Thunderbolt, the upstreamc4af8e3fecd03b0aedcd38145955605cfebe7e3a thunderbolt: Clear registers properly when auto clear isn't in use that landed in 6.3.3. One more quick test: could you check with 6.4-rc7 kernel in OBS Kernel:HEAD? If the problem is seen there, we'll need to report it to the upstream.
Meanwhile, I'm building a test kernel with the revert of the two recent TB patches (including the above mentioned one). It's being built in OBS home:tiwai:bsc1212587 repo. Please give it a try later once after the build finishes.
Kernel 6.4-rc7 is fixed. The test kernel in home:tiwai:bsc1212587 failed to build.
OK, at least, the bug will be fixed in the next week :) But it's still interesting whether those patches do right or not. I'm rebuilding the same repo, OBS home:tiwai:bsc1212587. Let's cross fingers. Also, another test build is found in OBS home:tiwai:bsc1212587-2 repo. This syncs the all changes in drivers/thunderbolt/*. Please try it later if you have time, too.
(In reply to Takashi Iwai from comment #11) > OK, at least, the bug will be fixed in the next week :) > > But it's still interesting whether those patches do right or not. > I'm rebuilding the same repo, OBS home:tiwai:bsc1212587. Let's cross > fingers. > > Also, another test build is found in OBS home:tiwai:bsc1212587-2 repo. This > syncs the all changes in drivers/thunderbolt/*. Please try it later if you > have time, too. Both kernels work correctly.
OK, thanks, it's good to know! I guess we need to backport relevant ones to SLE15-SP5, too (which had already the patch that caused the problem). As stable branch will be upgraded to 6.4 on the next Monday, I'll leave this. If 6.4-rc8 comes up and the upgrade is delayed, I'll merge the backport instead.
TW is switching to 6.4.x. Let's close.