• State
  • Locked Locked
  • Replies 2 replies
  • Subscribers 98 subscribers
  • Views 281 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TI EMAC Patch 33 Details

MikeP12806
Intellectual 450 points

I was wondering about getting some details on patch 33 to the 1.10 mvl kernel, ti_davinci_mv_1_0_emac_tcp_retransmit_fix_033.

It mentions a "ping from evm fails with no buffer space available error."  I was wondering if this was an intermittent error or a constant error.

If it is intermittent, what scenarios would the error come into play?

Also, I'm looking for some elaboration on the second description "when tcp is used with "connect" and the ack was not received, retransmission was only happening once"

What is the "connect" part of the statement?

 

It looks like in the path, the TX_INTERRUPT is getting enabled, where it was before disabled and many of the default configuration defines we removed.  I would be interested in understanding how this change solves this behaivor.

 

We have a custom board that is experiencing some issues that I feel may be related to not having this patch applied, but I would like to get some more information to verify my hunch.

 

Thanks.

  • 0
    TI__Expert 8250 points

    MikeP,

    Though i do not have the complete context, i do recollect the "Ping from evm Fails" scenario and the proposed patch to fix the issue.

     

    When you run flood ping from the EVM, the device will generate a stream of ECHO_REQUEST tokens without waiting for the corresponding response(unlike the case with normal ping/adaptive ping where the next packet is not generated until timeout/response received).  When the TX_INTERRUPT_DISABLE logic is enabled, the completed Tx buffer descriptors and corresponding skbuffs are not serviced/released until threshold is reached/Rx interrupt arrives. This proposed optimization doesnt fare well in scenario where there is burst of Tx traffic as in the case above. When all skbuffs are allocated, you will witness buffer allocation failure and the flood ping will start printing  stream of "EEEEE" indicating error condition. At the same time if you initiate Ping to the EVM from PC (which will trigger Rx interrupts) the flood ping will also recover.

    I do not entirely recollect if the situation is intermittent or unrecoverable

    Regards

    Sriram

  • 0 in reply to Srirama Govindarajan
    Intellectual 450 points

    Sriram,

    Thanks you for your response, it is very helpful.

    Do you or any of your colleagues have any information on the other aspect of the fix "when tcp is used with "connect" and the ack was not received, retransmission was only happening once"?

    Thanks,

    Mike