Faulty Ethernet connection prevented network CIFS mount at boot

[T_2]

I know this is a long post but I wanted to give back in the hope this helps others who might run across this weird problem. It has been plaguing us for many years. No where on the intertubes can I find anything like this. Maybe a Linux driver fix comes of this also.

We have 3 identical Dell Precision WorkStation 380 desktops each with a Broadcom Limited NetXtreme BCM5751 Gigabit Ethernet PCI Express network card and driver: tg3. All running up to date Linux Mint 19 MATE 64 bit. Although this problem has been present since Mint 17

They are part of a small network that has a mixture of other Linux machines, Window 7 Pro machines, an iMac, and 2 CentOS 6 servers. Because of the diversity of OS’s we set up Samba on one of the servers for file sharing and backup.

We shutdown our computers at the end of the day and power them back up in the morning. Servers power up first of course. As such we set all the desktops to reconnect to the network share at boot. They all mount the share just fine except for one of the Dell 380’s. Over the years I tried many things to try to get this system to behave. Even the fstab _netdev switch did not work. The only thing that would work is to set the problem child’s fstab to not mount the share at boot and set up a script in Mint’s Startup Applications to wait 60 seconds and then mount the share. It delays the startup but it seems to work.

Fast forward to few weeks ago I had some time to update this one machine from Mint 18.3 to 19 as the user was out of the office for a few weeks. First did a full backup and a nuke and pave install. I noticed the backup of that one machine takes this machine 3 times longer to backup over our network as the other 380’s. When I was setting up the system for all the users preferences as well as the share mount at boot I started to wonder again why does this machine behave different from the others especially since the 2 other identical machines do not have a boot share mount problem.

That’s when I ran inxi -F on one of the working 380’s and compared it to the output of same command on the problem system. As expected they were identical in all aspects except for one. The good 380 connects to the network at 1Gb but the problem system only at 100Mb. I then tried several things to force the network card to connect at 1Gb. Nothing worked. It could only connect at 100Mb. So I googled what is the difference between 100Mb and 1Gb and discovered the 100Mb only uses 4 of the 8 wires of an Ethernet cable where 1Gb uses all 8. Well time to check the cable. Found pin 8 was open – used by 1Gb but not by 100Mb. Quickly found the wire on a punchdown terminal block in the server room was in the right spot but it was never punched down so it was not making a connection. Its been that way since it was installed 12 years ago.

Once that was fixed I now had a 1Gb connection on the problem child! Yes! One problem fixed. A short time later I decided to look at the share mount at boot problem on this machine. So just for giggles changed it’s fstab file to match the other Linux machines and disabled the wait 60 seconds to mount script. Rebooted and SUCESS! The problem child was not longer a problem child. It mounted the share at boot just like the rest of the machines.

Now your assignment - if choose to accept it - is to tell me what causes this problem? Network card hardware or firmware? Is it something in the Linux software or drivers. Why should it care what speed the network is at if I use the fstab _netdev switch?