ixgbevf in dpdk doesn't seems to work (dpdk-stable-19.11.5)

ixgbevf in dpdk doesn't seems to work (dpdk-stable-19.11.5)

[Mohan Gyara]

[-- Attachment #1: Type: text/plain, Size: 1095 bytes --]

Hi Team,
In our application we use 2 interfaces one to communicate towards gNB and another to communicate with higher layers in a UE simulator NR stack. Now we use a PCI whose driver is i40e and is bind to dpdk vfio-pci driver and this works perfectly towards higher layer, but the other interface PCI whose is ixgbe driver is bind to vfio-pci dpdk driver and has some issues. 
The issues are:1. Application crashes when "rte_eth_dev_mac_addr_add" function is called when return with -1.  "on a 82599 VF, adding again the same MAC addr is not an idempotent operation..." this is a comment, but the function ixgbevf_add_mac_addr  return -1 when memcmp(hw->mac.perm_addr, mac_addr) == 0) is true. But above I commented the "return -1" and just returned from it, so crash is not seen.2. After above 1, application is started successfully but then I do not see any messages coming on this interface. This is the problem? 
Could you guide me what is the issue? I see only issue with ixgbvf but not for i40evf.
I am using dpdk-stable-19.11.5 version.

Kindly help me here.
Regards,Mohan

[-- Attachment #2: Type: text/html, Size: 3583 bytes --]

Re: ixgbevf in dpdk doesn't seems to work (dpdk-stable-19.11.5)

[Bruce Richardson]

+ ixgbe maintainers on CC

On Fri, Jun 03, 2022 at 08:14:00AM +0000, Mohan Gyara wrote:
>    Hi Team,
>    In our application we use 2 interfaces one to communicate towards gNB
>    and another to communicate with higher layers in a UE simulator NR
>    stack. Now we use a PCI whose driver is i40e and is bind to dpdk
>    vfio-pci driver and this works perfectly towards higher layer, but the
>    other interface PCI whose is ixgbe driver is bind to vfio-pci dpdk
>    driver and has some issues.
>    The issues are:
>    1. Application crashes when "rte_eth_dev_mac_addr_add" function is
>    called when return with -1.
>      "on a 82599 VF, adding again the same MAC addr is not an idempotent
>    operation..." this is a comment, but the function ixgbevf_add_mac_addr
>    return -1 when memcmp(hw->mac.perm_addr, mac_addr) == 0) is true.
>    But above I commented the "return -1" and just returned from it, so
>    crash is not seen.
>    2. After above 1, application is started successfully but then I do not
>    see any messages coming on this interface. This is the problem?
>    Could you guide me what is the issue? I see only issue with ixgbvf but
>    not for i40evf.
>    I am using dpdk-stable-19.11.5 version.
>    Kindly help me here.
>    Regards,
>    Mohan

Re: ixgbevf in dpdk doesn't seems to work (dpdk-stable-19.11.5)

[Mohan Gyara]

[-- Attachment #1: Type: text/plain, Size: 1621 bytes --]

 Hi,
I am blocked for almost 2 weeks now, I have browsed ixgbevf driver code but couldn't get any clue. kindly help me with this issue.

Regards,Mohan
    On Friday, 3 June, 2022, 02:29:58 pm IST, Bruce Richardson <bruce.richardson@intel.com> wrote:  
 
 + ixgbe maintainers on CC

On Fri, Jun 03, 2022 at 08:14:00AM +0000, Mohan Gyara wrote:
>    Hi Team,
>    In our application we use 2 interfaces one to communicate towards gNB
>    and another to communicate with higher layers in a UE simulator NR
>    stack. Now we use a PCI whose driver is i40e and is bind to dpdk
>    vfio-pci driver and this works perfectly towards higher layer, but the
>    other interface PCI whose is ixgbe driver is bind to vfio-pci dpdk
>    driver and has some issues.
>    The issues are:
>    1. Application crashes when "rte_eth_dev_mac_addr_add" function is
>    called when return with -1.
>      "on a 82599 VF, adding again the same MAC addr is not an idempotent
>    operation..." this is a comment, but the function ixgbevf_add_mac_addr
>    return -1 when memcmp(hw->mac.perm_addr, mac_addr) == 0) is true.
>    But above I commented the "return -1" and just returned from it, so
>    crash is not seen.
>    2. After above 1, application is started successfully but then I do not
>    see any messages coming on this interface. This is the problem?
>    Could you guide me what is the issue? I see only issue with ixgbvf but
>    not for i40evf.
>    I am using dpdk-stable-19.11.5 version.
>    Kindly help me here.
>    Regards,
>    Mohan
  

[-- Attachment #2: Type: text/html, Size: 3258 bytes --]

ixgbevf in dpdk doesn't seems to work (dpdk-stable-19.11.5)

[Mohan Gyara]

[-- Attachment #1: Type: text/plain, Size: 1096 bytes --]

Hi Team,
In our application we use 2 interfaces one to communicate towards gNB and another to communicate with higher layers in a UE simulator NR stack. Now we use a PCI whose driver is i40e and is bind to dpdk vfio-pci driver and this works perfectly towards higher layer, but the other interface PCI whose is ixgbe driver is bind to vfio-pci dpdk driver and has some issues. 
The issues are:1. Application crashes when "rte_eth_dev_mac_addr_add" function is called when return with -1.  "on a 82599 VF, adding again the same MAC addr is not an idempotent operation..." this is a comment, but the function ixgbevf_add_mac_addr  return -1 when memcmp(hw->mac.perm_addr, mac_addr) == 0) is true. But above I commented the "return -1" and just returned from it, so crash is not seen.2. After above 1, application is started successfully but then I do not see any messages coming on this interface. This is the problem? 
Could you guide me what is the issue? I see only issue with ixgbvf but not for i40evf.
I am using dpdk-stable-19.11.5 version.

Kindly help me here.
Regards,Mohan 

[-- Attachment #2: Type: text/html, Size: 2129 bytes --]