DP83826E: Link loss after successful autonegotiation completed

Andrea Tessadri

Part Number: DP83826E

I am using DP83826E with Linux SDK ti-processor-sdk-linux-rt-am335x-evm-07.03.00.005.

As soon as I connect the board to my laptop PC with a straight cable (so I expect the device to work in cross cable mode) I see the link going up and down.

I added some debug information to print the PHYSTS register.

Apparently, the device is trying every mode of operation starting from 10Base-T Half Duplex and then 100Base-T Full Duplex.

Even when the PHYSTS register detects signal and descrambler lock (0x4F15), the link will drop at next polling of the status register.

Which registers should I configure to have the PHY working in a cross cable mode ? (if I connect it to a switch it works!)

[ 54.483271] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x786D
[ 54.483518] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x4115
[ 54.483801] 000: prueth pruss_eth eth1: Link is Up - 100Mbps/Full - flow control off
[ 54.483822] 000: TI DP83826NC 4a332400.mdio:01: PHY state change NOLINK -> RUNNING
[ 54.483861] 000: IPv6: ADDRCONF(NETDEV_CHANGE): eth1: link becomes ready
[ 55.523375] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x7849
[ 55.523621] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x0912
[ 55.523661] 000: prueth pruss_eth eth1: Link is Down
[ 55.523672] 000: TI DP83826NC 4a332400.mdio:01: PHY state change RUNNING -> NOLINK
[ 56.563312] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x786D
[ 56.563563] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x0F15
[ 56.563845] 000: prueth pruss_eth eth1: Link is Up - 100Mbps/Full - flow control off
[ 56.563866] 000: TI DP83826NC 4a332400.mdio:01: PHY state change NOLINK -> RUNNING
[ 57.603273] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x7849
[ 57.603517] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x4912
[ 57.603560] 000: prueth pruss_eth eth1: Link is Down
[ 57.603570] 000: TI DP83826NC 4a332400.mdio:01: PHY state change RUNNING -> NOLINK
[ 58.643276] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x7849
[ 58.643523] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x0902
[ 59.683270] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x786D
[ 59.683518] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x4F15
[ 59.683799] 000: prueth pruss_eth eth1: Link is Up - 100Mbps/Full - flow control off
[ 59.683821] 000: TI DP83826NC 4a332400.mdio:01: PHY state change NOLINK -> RUNNING
[ 60.723388] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x7849
[ 60.723634] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x0912
[ 60.723674] 000: prueth pruss_eth eth1: Link is Down
[ 60.723686] 000: TI DP83826NC 4a332400.mdio:01: PHY state change RUNNING -> NOLINK
[ 61.763269] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x786D
[ 61.763514] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x4F15
[ 61.763796] 000: prueth pruss_eth eth1: Link is Up - 100Mbps/Full - flow control off
[ 61.763818] 000: TI DP83826NC 4a332400.mdio:01: PHY state change NOLINK -> RUNNING
[ 62.803317] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x7849
[ 62.803585] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x4912
[ 62.803626] 000: prueth pruss_eth eth1: Link is Down
[ 62.803637] 000: TI DP83826NC 4a332400.mdio:01: PHY state change RUNNING -> NOLINK
[ 63.843269] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x7849
[ 63.843515] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x0902
[ 64.883273] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x7849
[ 64.883520] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x0912

I add register values when connected to an ethernet switch until I disconnect the cable:

[ 61.683376] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x786D
[ 61.683623] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x5115
[ 61.683908] 000: prueth pruss_eth eth1: Link is Up - 100Mbps/Full - flow control off
[ 61.683930] 000: TI DP83826NC 4a332400.mdio:01: PHY state change NOLINK -> RUNNING
[ 61.683971] 000: IPv6: ADDRCONF(NETDEV_CHANGE): eth1: link becomes ready
[ 62.723894] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x786D
[ 62.724149] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x5715
[ 63.763383] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x786D
[ 63.763628] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x5715

....................

[ 76.243409] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x7849
[ 76.243678] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x5912
[ 76.243722] 000: prueth pruss_eth eth1: Link is Down
[ 76.243732] 000: TI DP83826NC 4a332400.mdio:01: PHY state change RUNNING -> NOLINK
[ 77.283380] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x7849
[ 77.283629] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x5902

over 4 years ago

0 Lucas Wolter over 4 years ago

TI__Mastermind 27800 points

Hi Andrea,

Can you provide a schematic for review? You can try forcing the speed MDI/MDIX through register 0x0. When forcing the speed in MDI/MDIX resolution is this issue still seen?

Thanks,

Lucas

All information in this correspondence and in any related correspondence is provided “AS IS” and “with all faults”, and it is subject to TI’s Important Notice (www.ti.com/.../important-notice.shtml).

0 Andrea Tessadri over 4 years ago in reply to Lucas Wolter

Intellectual 567 points

Here is our schema.

If I force the speed to 100 Mbps and Full Duplex (register 0x0 at 0x2100) I am able to get the link up also when the cable is connected to my PC.

We are also wandering if straight/cross cable detection is supported in basic mode.

Looking forward

Andrea

0 Lucas Wolter over 4 years ago in reply to Andrea Tessadri

TI__Mastermind 27800 points

Hi Andrea,

Yes, auto-MDIX feature should be fully supported by the device. What is the cable length?

Are the fast-link features enabled (register 0xB)? If so, does this link drop occur when fast-link drop is disabled?

We're reviewing your schematic and will provide feedback by Friday.

Thanks,

Lucas

All information in this correspondence and in any related correspondence is provided “as is” and “with all faults”, and it is subject to TI’s Important Notice (http://www.ti.com/corp/docs/legal/important-notice.shtml).

0 Andrea Tessadri over 4 years ago in reply to Lucas Wolter

Intellectual 567 points

Hi Lucas,

here is the registers status with the link up but without any transmission (even if I am pinging).

So the fast-link drop isn't enabled.

The cable is 2 m long.

[ 173.927654] 000: TI DP83826NC 4a332400.mdio:01: BMCR [0x00] = 0x2100
[ 173.927921] 000: TI DP83826NC 4a332400.mdio:01: CR1 [0x09] = 0x0000
[ 173.928166] 000: TI DP83826NC 4a332400.mdio:01: CR2 [0x0A] = 0x0100
[ 173.928410] 000: TI DP83826NC 4a332400.mdio:01: CR3 [0x0B] = 0x0000
[ 173.928655] 000: TI DP83826NC 4a332400.mdio:01: BMSR [0x01] = 0x784D
[ 173.928898] 000: TI DP83826NC 4a332400.mdio:01: PHYSTS [0x10] = 0x0605

0 Lucas Wolter over 4 years ago in reply to Andrea Tessadri

TI__Mastermind 27800 points

Hi Andrea,

We're looking into your question and will get back to you by Friday.

Thanks,

Lucas

0 Andrea Tessadri over 4 years ago in reply to Lucas Wolter

Intellectual 567 points

Hi Lucas,

I add also the part of the schematic for the RJ45 connector.

0 Lucas Wolter over 4 years ago in reply to Andrea Tessadri

TI__Mastermind 27800 points

Hi Andrea,

We'll take a look at this too and get back to you soon.

Thanks,

Lucas

0 Lucas Wolter over 4 years ago in reply to Lucas Wolter

TI__Mastermind 27800 points

Hi Andrea,

I didn't see any major issues with the schematic. Can you confirm the values of registers 0x467 and 0x468 in both working and non-working cases?

Can you also provide values for registers 0x11, 0x18, 0x19, 0x302, 0x303, 0x304, 0x305, 0x306, and 0x460?

Thanks,

Lucas

0 Andrea Tessadri over 4 years ago in reply to Lucas Wolter

Intellectual 567 points

Hi Lucas,

we have been able to fix autonegotiation, both with straight cable (switch connection) and cross link (PC connection). It was probably due to a RJ45 connector not well soldered since we could't see TX pulses at the beginning.
Later on we discovered also that a pull-up resistor on pin 28, in another part of the circuit, forced the device to RMII mode instead of MII mode expected by PRU (we had a 50 MHz CLK instead of 25 MHz).
We removed the pull up and we added a pull down to force MII mode.

It works now.

Thank you as well for having revised our schema.

Andrea

Interface

Interface forum

DP83826E: Link loss after successful autonegotiation completed