[lldp-devel] [PATCH 1/2] ecp fix endless loop issue
john.fastabend at gmail.com
Fri Mar 22 04:46:57 UTC 2013
On 03/12/2013 03:59 AM, Thomas Richter wrote:
> The ECP protocol calls the ECP_TX_INIT_TRANSMIT
> in an endless loop under certain conditions. The bug appears on
> RHEL 6.3 and RHEL 6.4 (probably on RHEL 6.2 too), but not on
> Fedora 18. I have not tested any other distribution.
> The problem shows up when a link is down *and* has QBG enabled in
> the lldpad configuration file.
> When you start lldpad, you get endless messages of:
> 16:13:34.979858 ecp_tx_run_sm-eth1: ecp_tx - ECP_TX_INIT_TRANSMIT
> 16:13:34.979861 ecp_somethingChangedLocal:eth1 vd->ecp.tx.localChange to true
> 16:13:34.980938 ecp_localchange_handler:eth1 ecp.tx.localChange 1
> 16:13:34.980942 ecp_tx_run_sm-eth1: ecp_tx - ECP_TX_INIT_TRANSMIT
> 16:13:34.980945 ecp_somethingChangedLocal:eth1 vd->ecp.tx.localChange to true
> 16:13:34.982013 ecp_localchange_handler:eth1 ecp.tx.localChange 1
> Those messages appear on the screen with -V7 option only, but in an
> extremely high frequency. Without this option you do not see the
> messages, but the problem is still there.
> The basic problem is that the VDP module is started and triggers
> ECP communication. But when the link is down the function
> ecp_tx_run_sm() does not advance to the next state. Instead if
> keeps going for ever in state TX_INIT_TRANSMIT.
> To remedy this situation, remove the check for link active. It also
> changes the portEnabled variable to false when the link is not
> active. The change of the portEnable variable to false prevents
> the state machine from advancing to the next state.
> Reported-by: Mijo Safradin <mijo at linux.vnet.ibm.com>
> Signed-off-by: Thomas Richter <tmricht at linux.vnet.ibm.com>
Both applied thanks.
More information about the lldp-devel