On Wed, Mar 24, 2010 at 02:34:40AM -0700, Ahmed A wrote: > Hello, > > I am trying to setup a Linux system to snoop FCoE frames. The Linux RedHat > 5.4 system has a dual port 10G QLogic CNA(hba), that are configure as regular > NICs. I configured a bridge between the these two 10G ports, using the > built in bridge code in the kernel, and using bridge-utils package. I then > connected one of the bridge ports to an 10G FCoE Initiator in another Linux > server and the other 10G bridge port to a Nexus switch. For some reason the > FCoE Initiator is not logging into the Nexus switch, which it does if > directly connected. Am I missing something basic? Is there any > handshake/negotiation that happens below the ethernet level between the > Initiator port and Nexus port? > > Using the same setup, I am able to snoop iSCSI packets between an iSCSI > Initiator and iSCSI Target. So, the bridging and snooping (using ethereal) > does work. > > > > Linux RedHat 5.4 > (Dual port 10G CNA, configured part of a > bridge). > | | > | | > FCoE Initiator ______| |______Nexus Switch >
VLAN stripping is one possibility. Another is DCB configuration. DCB parameters are exchanged via LLDP, which uses a link local multicast address. I don't think the bridging code will forward LLDP frames. If the initiator is configured with DCB_REQUIRED="yes", then FCoE traffic will never start if DCB cannot be configured. You can try setting that to "no", and depending on the switch configuration things may proceed. But running FCoE without some reliable flow control is asking for I/O errors, and from what I know all vendors are requiring DCB for a supported configuration. - Chris _______________________________________________ devel mailing list [email protected] http://www.open-fcoe.org/mailman/listinfo/devel
