PCI Interrupt Routing kernel patch for Jetson 5.0 DP (34.1)

@vidyas @Manikanta

The Jetson 5.0 DP kernel has the same issue of not correctly assigning interrupts in certain situations. This post shows how to solve it for previous versions, but the file structure of the kernel source is slightly different and uses much different code. Xavier not routing PCI interrupts across PEX8112 bridge - #22 by vidyas

Here’s another thread about the interrupt issue, provided so the original author (@john.mcinnis) can see the fix too: Default IRQ of 0

Can you please post the fix for 5.0 DP?

Can anyone from NVIDIA help? This seems to be an issue with a handful of PCIe cards.

Is anyone at NVIDIA working on this? Is there a plan to have a fix provided for this?

@vidyas @Manikanta,

Would one of you (or anyone at NVIDIA) please let me know if this is something that is either being worked on, scheduled to be worked on, or not slated to be worked on? I need to know if I should roll back to JP 4.6.1 if there is no fix that will be delivered soon.

I’d appreciate at least a quick response on NVIDIA’s position on getting this fix into JP 5.0 (and beyond). Thank you.

Hi, what is the exact issue you are observing with JP 5.0?

It’s the same issue that JP 4.6 had where interrupts aren’t being assigned correctly through bridge devices. In my first post on this thread I reference an older thread (Xavier not routing PCI interrupts across PEX8112 bridge - #22 by vidyas) that you provided the fix for, which involved editing the kernel code, recompiling, and flashing. This worked in 4.6 and 4.6.1, but the 5.0 DP files are different and don’t even use some of those functions.

So I’m asking for whatever needs to be modified in the 5.0 kernel code to fix this interrupt assignment issue.

The Designware code has changed quite a bit. Give me a couple of days for me to figure out what change is needed to get legacy interrupt routing fixed in K5.10.

Thank you.

Hi @chris.brahmer
I tried connecting a USB 3.0 device behind a PCIe switch and disabled MSI interrupts in the system (by adding ‘pci=nomsi’ to the kernel command line) and it is working fine.
I’m attaching the full log here.

legacy_intr.txt (115.4 KB)

I’m using the BSP and Sample Root Filesystem from link

Could you please help me understand better what exactly is the issue you are observing?
Probably boot log along with the output of ‘sudo lspci -vvvv’ and ‘cat /proc/interrupts | grep -i pci’ are good starting points.

Thanks for the reply. I’ll try the kernel command line update you suggested using 5.0 DP, but I can’t test that for a while as our project deadline is looming and I’ve had to forge ahead with 4.6.1.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.