[dpdk-dev] Kernel Crash bug of ixgbevf kernel module in "Intel(R) 10GbE PCI Express Virtual Function Driver Version: 4.0.3 Release: 1"

[Sam]

I don't know if it's because ixgbevf driver call "ixgbevf_close()" by
"dev_close()" without rtnl_lock().

Refer to log "bond1: Releasing active interface enp1s16f1", bond is
releasing its slaves, then call "dev_close()" function without rtnl_lock().
At the same time, "ixgbevf_service_task()" loop to
"ixgbevf_reinit_locked()" function. Then two thread operate same address,
then bug happens.

2018-01-31 15:44 GMT+08:00 Sam <batmanustc at gmail.com>:

> wrong order, actually is this:
> 4. stop bond1(ifdown bond1), stop dpdkb2(rte_eth_dev_stop), sleep 5 second,
> start dpdkb2(rte_eth_dev_start), start bond1(ifup bond1).
>
> 2018-01-31 15:34 GMT+08:00 Sam <batmanustc at gmail.com>:
>
>> Hi all,
>>
>> There is a kernel crash bug of ixgbevf kernel module in "Intel(R) 10GbE
>> PCI Express Virtual Function Driver Version: 4.0.3 Release: 1"
>>
>> How to produce:
>> 1. Use SRIOV, like this:
>> sudo /usr/local/share/openvswitch/scripts/dpdk_nic_bind --status
>>
>> Network devices using DPDK-compatible driver
>> ============================================
>> 0000:01:00.0 'Ethernet Controller 10-Gigabit X540-AT2' drv=igb_uio
>> unused=ixgbe
>> 0000:01:00.1 'Ethernet Controller 10-Gigabit X540-AT2' drv=igb_uio
>> unused=ixgbe
>>
>> Network devices using kernel driver
>> ===================================
>> 0000:01:10.0 'X540 Ethernet Controller Virtual Function' if=enp1s16
>> drv=ixgbevf unused=bak,igb_uio
>> 0000:01:10.1 'X540 Ethernet Controller Virtual Function' if=enp1s16f1
>> drv=ixgbevf unused=bak,igb_uio
>> 0000:08:00.0 'I350 Gigabit Network Connection' if=eth2 drv=igb
>> unused=igb_uio
>> 0000:08:00.1 'I350 Gigabit Network Connection' if=eth3 drv=igb
>> unused=igb_uio
>>
>> Other network devices
>> =====================
>> <none>
>>
>> 2. bond enp1s16 and enp1s16f1 into bond1, by /etc/sysconfig/ifcfg-bond1.
>> 3. bond 0000:01:00.0 and 0000:01:00.1 in ovs-dpdk into dpdkb2, by dpdk
>> api.
>> 4. stop dpdkb2(rte_eth_dev_stop), stop bond1(ifdown bond1), sleep 5
>> second, start bond1(ifup bond1), start dpdkb2(rte_eth_dev_start).
>>
>> After several times, bug happens, attachment is vmcore-dmesg.txt
>>
>
>