Version 6.11.5 -> 6.12.6 (also happened with 6.12.5)

 

Hardware:

Intel® Core™ i5-4670S

ASUS H87M-E Motherboard

Broadcom NetXtreme BCM5751 Gigabit Ethernet

I can provide more details here if needed but I'm not sure it's relevant.

 

I'm trying to upgrade from 6.11.5 to 6.12.x since I am out of date and things are starting to become incompatible.  When I run the upgrade, on reboot my broadcom NIC is no longer detected and my network doesn't come up.  I ran into this a few months ago with 6.12.5 but didn't have time to debug it then, but it's also occurring with 6.12.6.  The error I see in the boot prompt is "Device 'eth0' not found" or something similar.

 

I've been doing some digging and troubleshooting myself, and it looks like the tg3 module is no longer being loaded in 6.12.6 even though it's still present.  Using lspci when the system is on 6.11.5 I can see the card listed and that module is listed with it, but when I try to run 6.12.6, the module doesn't get loaded.  I've tried loading it manually with modprobe, but the card still doesn't seem to be detected, and I've also tried modifying the /config/go file to load the module in there as well, but it still didn't detect the card.  It's possible that there are maybe parameters that are needed to be given to the module but I'm not sure what they are.

 

I've attached a diagnostics bundle from one of the boot ups in 6.12.6 when the card wasn't detected.  If there's additional info that's needed, please let me know.

 

The problem is repeatable and I've been switching versions back and forth during debugging with copies I've made of the USB drive.  Based on the fact that the card has worked for about a year and a half and continues to work when I revert to 6.11.5, to me that points to a configuration issue.  I've hit the limit of what I'm able to figure out at this point so any additional help here would be appreciated. 

 

Thanks.

hyperion-diagnostics-20231228-1610.zip

Just to confirm that NIC is in addition to this one correct?

 

    

Subsystem: ASUSTeK Computer Inc. H81M-C Motherboard [1043:8554]
    Kernel driver in use: r8169
    Kernel modules: r8169

 

No other NIC is being detected, this is independent of a driver/module being loaded or not, please post diags from v6.11

Yes that's correct.  The one that is showing up there is the onboard NIC that I usually have disabled; it's a realtek 8111g chip.  I had it turned on at one point when I was trying different changes to see if I could get something to work.  

 

Here are the diagnostics from 6.11.  Thanks for taking a look, I appreciate it.

hyperion-diagnostics-20231229-1554.zip

03:00.0 Ethernet controller [0200]: Broadcom Inc. and subsidiaries NetXtreme BCM5751 Gigabit Ethernet PCI Express [14e4:1677] (rev 01)
    Subsystem: Broadcom Inc. and subsidiaries NetXtreme BCM5751 Gigabit Ethernet PCI Express [14e4:1677]
    Kernel driver in use: tg3
    Kernel modules: tg3

 

Yep, here it is, no idea why is not being detected with v6.12, and like mentioned this is not driver dependent, the NIC should always show up in the PCI devices even if no driver is being loaded, and of course it will never work if it's not being detected.

 

Did you try a different PCIe slot? Also look for a BIOS update and/or reset CMOS.

 

Alright so I started with trying a different PCIe slot as suggested.  The board has 4 slots, a x16 slot at the top and three x1 slots underneath it.  Originally it was in the 1st x1 slot.  I tried the second x1 slot and it still didn't work.  I then tried the x16 slot and then it worked.  I also tried it in the bottom x1 slot and it didn't work there either.  So it appears that the card only works in the x16 slot on 6.12.  I can't find a block diagram to confirm for this motherboard, but maybe this may be due to the x16 slot typically going directly to the cpu instead of through the chipset?  Is it possible that the chipset driver configuration changed or support for this one was removed?

 

As another data point, the NIC has a PXE boot process that it tries to run when the computer first boots before the OS is loaded.  In all slots this process ran, which to me indicates that the bios recognized the card.

 

I can run the card in the top slot so that it works, but it seems like there's something that changed that has caused this issue to occur with the other slots.

Here's a diagnostics zip from a boot with the card in the x16 slot, in case that's of value.

hyperion-diagnostics-20231230-1003.zip

This is very strange, maybe the PCH slots are not being correctly initialized with the newer kernel, but don't remember ever seeing that before, do you have any other card you could connect there to see if it's also not detected?

Yea I have an old wifi x1 card I can test with.  So I tested the card first in the x16 slot and then in the top x1 slot, both with 6.12, and in both cases, the card appeared to be detected, as it was listed by lspci.  I took diagnostics of both boots.  So it appears the other x1 lanes can detect hardware in those slots, but there is an issue with the broadcom card for some reason.  I didn't have time today to do more investigation than that.

hyperion-diagnostics-20231231-0940_wifi_x16.zip hyperion-diagnostics-20231231-0946_wifi_x1.zip

Sorry, don't have more ideas other than use a different NIC (or board) is possible.