First, is your “uname -r” “4.4.38”? Or is it “4.4.38-tegra”? If not the latter, then the kernel has been changed and the modules need a new location (any module based feature would be missing if you didn’t rebuild all modules).
Second, alternate carrier board demand a different board support package…mainly a device tree difference.
Next, if you used a clone of any unit for flash we’ll need to know. In part because a clone used during flash must come from the exact same L4T release as it was originally created for (so for example you can’t flash an R28.1 clone into an R28.2 release). In some cases a clone may also hard code network setup related to the MAC address of the original unit, and thus a clone which does this (most clones don’t) would need the MAC address edited in “/etc”. Was any cloning involved?
Btw, it will be easier to read logs if you edit the post (the “pencil” icon shows up for editing if you hover your mouse over the quote icon in the upper right), highlight the log or output, then click on the “code” icon (looks like “</>”).
I ask about grabserial because some people mistakenly use the serial console port for a free UART.
About Current Settings…
This is not valid in route:
10.3.0.0 * 255.255.0.0 U 0 0 0 eth0
10.3.0.0 * 255.255.0.0 U 0 0 0 eth1
You have two network cards with the same subnet and same netmask. The two cards and their drivers will fight for which one should be used. I would expect that somewhere in the middle of any working connection that the drivers might suddenly get mixed up and traffic might get split across two NICs in such a way that it locks up.
Looking at ifconfig, this conflict is verified, and just gets worse:
eth0 Link encap:Ethernet HWaddr 00:04:4b:c0:5b:2e
inet addr:<u><i><b>10.3.1.1</b></i></u> Bcast:10.3.255.255 Mask:255.255.0.0
...
eth1 Link encap:Ethernet HWaddr 00:0c:8b:90:09:fe
inet addr:<u><i><b>10.3.2.1</b></i></u> Bcast:10.3.255.255 Mask:255.255.0.0
Not only do their subnets and netmasks collide, the two are the same IP address.
The two NICs on a single host require separate addresses (and in fact no two NICs on any network should have the same address), and their combination of subnet and netmask must be unique (route won’t behave correctly when the two overlap). Until traffic is generated such that both drivers try to deal with the traffic you won’t see a failure…as soon as there is traffic with a conflict the system will fail.