How does Ethernet driver know if the cable is plugged in?

Question

How does the /sys/class/net/ethx/carrier field get set?

I am trying to understand this field as it determines if a network cable is plugged in, but I am curious how it knows. This is specifically in regards to 10Gbase-KR and 1Gbase-KX.

I have a little knowledge in regards to CMSA/CD, but that is only for half duplex and NPL - pulses are for auto negotiation.

Is something going across the medium when no traffic is being sent?

I believe it may have to do with IDLE symbols, but I do not fully understand them.

Answer 1

Link detection is specific to the physical layer.

• The original coax Ethernet did not really have link state detection (some PHYs had a mechanism for detecting termination during transmission but it only worked when the device was actually trying to transmit something)
• 10BASE-T uses "link state pulses" which are sent when the line is idle to monitor link status.
• Twisted pair auto-negotiation uses "Fast link pulse bursts" to both determine the state of the link and to exchange auto-negotiation information. Once a link type has been negotiated link type specific mechanisms are used to monitor link status.
• Faster Ethernet standards generally transmit idle codes continuously.

If you want the full details of what a particular physical layer sends best to check the actual 802.3 standard.

Answer 2

Basically, auto negotiation tells the port which PHY to select. The PHY starts up and its PCS will then receive the carrier (or carriers for multilane), depending on the PHY flavor.

Answer 3

You can judge it by the LEDS signal light. If laptop says UNPLUGGED then you don't have a connection, and this can also be verified by the tinny LED by the RJ45. So OK, some laptops may not have LEDs but certainly the RJ45 by the switch port have LEDs?

Surely, there is ANOTHER device you can try. TV? DVD player with network? SOMETHING. RJ45 to RJ45, the LEDs should light up.

Hope this would help!