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 Thinks Resolved

DP83867CS: Kernel version

Prodigy 490 points

Replies: 10

Views: 298

Part Number: DP83867CS

Our customer is using our DP83867CS in their application. They found that the code on our Internet is for Kernel 4.2. However, they are now using the Kernel 3.18.20 version and it cannot be used.

Please advise if we can support the code for Kernel 3.18.20 version. If no, how can we modify the existing code for their application?

  • Hi David,

    I have started the process to check if we can support 3.18 kernel version driver. I will let you know what I find.

    -Regards
    Aniruddha
  • In reply to Aniruddha Khadye:

    Hi David,

    I could get the 3.18 Kernel Linux Driver and its attached with this reply.


    -Regards

    Aniruddha

    DP83867_Linux318_backport.zip

  • In reply to Aniruddha Khadye:

    Hi Genius ,
    PHY Ping failed after 3.18 kernel transplantation of dp83867cs

    PC side, network local connection settings, network configuration changes, Speed & Duplex for 100 Mbps, can ping. (default is automatic negotiation, automatic negotiation Ping is not available)

    Suspicion is a problem of automatic network negotiation. Do the following

    Hardware:

    Two pairs of twisted pairs A and B are connected to magnetics, using SGMII four-wire, RX_CTRL bundling mode 3, LED_0 bundling mode 2.

    Software:

    Register (BMCR), Address 0x0000 RW bit 12 = 1, Val = 0x1140

    Register (BMSR), Address 0x0001 R bit 5, Val = 7949

    Register (PHYCR), Address 0x0010 RW bit 11 = 1, Val = 0x5848

    Register (CFG2), Address 0x0014 RW bit 7 = 1, Val = 0x29c7

    Register (SGMII_ANEG_STS), address 0x0037 R bit 0, Val = 0x40

    After these attempts, it is still impossible to negotiate automatically.

    Please have a look. Thank you.
  • In reply to TianSen xue:

    Hi,

    I am able to fully understand your setup.

    "Two pairs of twisted pairs A and B are connected to magnetics, using SGMII four-wire, RX_CTRL bundling mode 3, LED_0 bundling mode 2."

    Kindly share the block diagram of your test setup and schematics of the DP83867 .

    Regards,
    Geet
  • In reply to Geet Modi:

    DP83867CS_SCH.pdfphy-debug (2).txt

     See attachments txt and PDF

    Thanks

    请查看附件txt和pdf

    See attachments txt and PDF

  • In reply to Geet Modi:

    Hi,

    See attachments txt and PDF

    Thanks

    3240.DP83867CS_SCH.pdf6663.phy-debug (2).txt

  • In reply to TianSen xue:

    Hi,

    The schematics shows the only two channels are used in MDI, Channel A and B. This is ok for 10/100Mbps. But from register 0x09, it looks like the PHY is advertising 1Gbps. 1Gbps cannot work on only two channels. Can you try disabling 1Gbps advertisement by writing 0x0000 to register 0x09? What is the link partner in this case? If it also the same DP83867 system, then please disable 1Gbps advertisement on the link partner as well.

    -Regards
    Aniruddha
  • In reply to Aniruddha Khadye:

    Link partners are PC-side, PC-side local connections, the default configuration is automatic negotiation (now the phenomenon is: if the PC-side local connection, the default configuration is changed to 100M or 10M, can ping pass; using default configuration, automatic negotiation can not Ping pass)

    Now the problem is that if you don't change the default configuration of the PC side, two channels are still used in Channel A and B (because our circuit only keeps four-track lines), and the link partner is the PC side (PC side keeps default settings automatically negotiated).

    Is there any way to change registers or circuit (keep 4 traces, channel A and B still use the premise of two channels), to achieve normal network connection (ping pass)

    Similar cases were found on E2E platform and closed without answering how to solve them. The links are as follows: e2e.ti.com/.../676515 Tisearch=e2e-sitesearch&keymatch=DP83867%20Reference%20Clock.

    Please help me analyze it.

    Thank you!

  • In reply to Aniruddha Khadye:

    HI,Khadye

    Please help me to see.

    Does dp83867 support four-line self-negotiation?

    How to set it up to meet the four-line self-negotiation? Is it okay to change registers? Or can changing the four-wire hardware circuit satisfy self-negotiation?

    If dp83867 does not support four-line self-negotiation, then we can only consider eight-line self-negotiation, but eight-line self-negotiation is not what we want.

    Thank you!
  • In reply to TianSen xue:

    I'm sorry,

    It should be auto-negotiation, not Self-negotiation.

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.