Reloading igb_avb.ko crashes the OS

[RikusW]

Doing insmod rmmod and then insmod again will crash the system entirely requiring a reboot. I used kernel 3.19.0-15-generic #15-Ubuntu SMP 64 bit.

[MantisClone]

I'm working with two intel x86_64 PCs running CentOS7 with two intel I210 cards installed, and I don't run into this issue. When I run the run_igb.sh script with an added line to rmmod igb_avb, the PC does not crash.

rmmod igb
rmmod igb_avb # this is the line I added
modprobe i2c_algo_bit
modprobe dca
modprobe ptp
insmod kmod/igb/igb_avb.ko

[eric-mann]

Not much meat to chew on with this bug report. Backtrace?

You can get similar bad behavior if you fail to unload the igb driver first. Two drivers operating on the same hw at the same time has 'undetermined behavior'.

[RikusW]

It only happened when I did "rmmod igb" first, when I left it in there was no crash. I'll try again with the above script.

I did a git pull a few minutes ago. It still crashes, Numlock doesn't even work after insmod kmod/igb/igb_avb.ko

[RikusW]

For now I'll just avoid doing rmmod, then all is fine or so it seems. I''m working for Uman an AVnu member company and will be working with Open AVB.

[RikusW]

This is the required commands to make it crash, if igb is left in there is no crash.

rmmod igb insmod kmod/igb/igb_avb.ko rmmod igb_avb insmod kmod/igb/igb_avb.ko

Totally dead after the last command. There is only one i210 in the PC.

[eric-mann]

You definitely do not want to leave igb loaded.

Can you define ‘totally dead’?

ekm

[RikusW]

Ironically leaving in igb prevents the crash, maybe because it prevents allocation of some resource.

The mouse and keyboard is unresponsive, numlock doesn't even change the led. I had to unplug the PC to reboot since the reset button is not connected. I haven't tried sysrq since I don't think its enabled.

[eric-mann]

So is igb_avb already loaded when you try to load it again?

Can you include the output of ‘lsmod’ before you run the script?

ekm

[RikusW]

Only igb is loaded beforehand igb_avb is not. Removing igb_avb and then reloading it crashes the PC only if igb is NOT loaded. I'll do lsmod when I'm back at work.

So running the script that DMats provided twice will cause the crash.

[RikusW]

$ lsmod Module Size Used by pci_stub 16384 1 vboxpci 24576 0 vboxnetadp 28672 0 vboxnetflt 28672 0 vboxdrv 425984 3 vboxnetadp,vboxnetflt,vboxpci cfg80211 548864 0 binfmt_misc 20480 1 snd_hda_codec_realtek 86016 1 snd_hda_codec_generic 69632 1 snd_hda_codec_realtek snd_hda_intel 36864 3 snd_hda_controller 32768 1 snd_hda_intel snd_hda_codec 143360 4 snd_hda_codec_realtek,snd_hda_codec_generic,snd_hda_intel,snd_hda_controller snd_usb_audio 184320 2 gpio_ich 16384 0 snd_usbmidi_lib 36864 1 snd_usb_audio snd_hwdep 20480 2 snd_usb_audio,snd_hda_codec intel_powerclamp 20480 0 snd_pcm 106496 4 snd_usb_audio,snd_hda_codec,snd_hda_intel,snd_hda_controller snd_seq_midi 16384 0 snd_seq_midi_event 16384 1 snd_seq_midi snd_rawmidi 32768 2 snd_usbmidi_lib,snd_seq_midi coretemp 16384 0 i915 1060864 4 kvm_intel 151552 0 kvm 487424 1 kvm_intel snd_seq 65536 2 snd_seq_midi_event,snd_seq_midi snd_seq_device 16384 3 snd_seq,snd_rawmidi,snd_seq_midi serio_raw 16384 0 mei_me 20480 0 video 20480 1 i915 snd_timer 32768 2 snd_pcm,snd_seq mei 90112 1 mei_me drm_kms_helper 131072 1 i915 snd 90112 22 snd_hda_codec_realtek,snd_usb_audio,snd_hwdep,snd_timer,snd_pcm,snd_seq,snd_rawmidi,snd_hda_codec_generic,snd_usbmidi_lib,snd_hda_codec,snd_hda_intel,snd_seq_device drm 348160 6 i915,drm_kms_helper lpc_ich 24576 0 shpchp 40960 0 soundcore 16384 2 snd,snd_hda_codec mac_hid 16384 0 parport_pc 32768 0 ppdev 20480 0 lp 20480 0 parport 45056 3 lp,ppdev,parport_pc autofs4 40960 2 hid_generic 16384 0 usbhid 53248 0 hid 110592 2 hid_generic,usbhid ahci 36864 2 libahci 32768 1 ahci igb 188416 0 i2c_algo_bit 16384 2 igb,i915 r8169 81920 0 dca 16384 1 igb ptp 20480 1 igb mii 16384 1 r8169 pps_core 20480 1 ptp