Have you modified +/etc/udev/rules.d/40-multipath.rules+ ?
What does the kernel say when discovering these EMC LUNs (prior to multipath loading)?
What I/O fabric layer are you using? Fibre? Ethernet? SRP? And it would seem that EMC is being used (from the hwhandler output) - is Powerpath loaded and in use?
Multipath does not do the device discovery and enumeration. That is done prior when the kernel boots and the I/O fabric layer driver is loaded. Multipath attempts to determine what the I/O paths are, which devices refer to the same WWID, and then (if whitelisted) map these to a single logical multipath device.
So if multipath runs in a device error of sorts (as per your kernel log) - that error very likely comes from deeper down the driver stack.
Can you flush and rerun multipath and post the output? E.g.
// devices may not be in use
[root@reuxeuls003 ~]# multipath -F
// run multipath discovery with verbosity at level 3
[root@reuxeuls003 ~]# multipath -v3
Well, then you need to go line-by-line through that verbose output and see what the errors multipath runs into, trying to use the problem I/O paths (scsi devices).
Or open a SR (service request) with Oracle to investigate the error - assuming you have an OL support agreement and a CSI number.