5

What happens if an error is detected at the lowest possible layer of the stack ISO/OSI for Ethernet?

For example, considering 100BASE-TX, it is implemented with a 4B5B coding this means that there are 16 combinations of bits are invalid. What happens if a receiver "reads" from the physical medium an invalid combination of bit, e.g. "00001"?

  • It discards that bit (invalidating the entire frame)
  • It select the "most similar" combination
  • or something else?

I know that upper layers can detect the error, e.g. the MAC layer via the FCS field, but I don't know how the transreceiver works with an error detected in the signal.

3
  • As described in en.wikipedia.org/wiki/4B5B 4B5C don't provide error correction, so 2-nd doesn't look like option.
    – mmv-ru
    Feb 20, 2017 at 21:07
  • @mmv-ru you are right. But it seems strange to me that the physical layer simply discards the invalid bits, in that case it would provide a frame with an incorrect length to mac layer.
    – ocirocir
    Feb 20, 2017 at 22:01
  • The physical interface will not pass on a damaged frame to the data-link layer. It discards the damaged frame, and it increases the input error count on the interface.
    – Ron Maupin
    Feb 20, 2017 at 22:26

1 Answer 1

3

For the specific case of 100BASE-TX and the 100Mbps MII interface: there is a receiver signal RX_ER.

The Physical Coding Sublayer (PCS) specification says:

24.2.2.1.7 Invalid code-groups

...The PCS indicates the reception of an Invalid code-group on the MII through the use of the RX_ER signal.

The Reconciliation Sublayer and Media Independent Interface specification says:

22.2.1.5 Response to RX_ER indication from MII

If, during frame reception, both RX_DV and RX_ER are asserted, the Reconciliation sublayer shall ensure that the MAC will detect a FrameCheckError in that frame.

So the PHY part indicates the error out of band, and the MAC does what it likes, as long as it ends up generating a FrameCheckError.

2
  • Thank you, but what happens to the invalid bit group? Will it be ignored (and consequently the frame delivered to MAC will be shorter than expected)?
    – ocirocir
    Feb 22, 2017 at 7:18
  • @RicoRico The interface is synchronous, so there won't be a gap. The 4 bits of RX data on that clock cycle are undefined.
    – richardb
    Feb 22, 2017 at 18:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.