Mac 00:00:00:00:00:01

I installed SDKmanager on a virtual machine.

SDKmanager gets stuck flashing the board. Last command in log is:
tegrarcm --chip 0x21 0 --updatesig rcm_list_signed.xml

Unfortunately, sdkmanager does not support on VM.
Please install dual boot on your host or find another ubuntu host for it.

Ok, I set up a fresh Ubuntu 18.04 PC. Got Jetson Nano flashed to SDK 4.4.

Still same issue.

3: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP group default qlen 1000
link/ether 00:00:00:00:00:01 brd ff:ff:ff:ff:ff:ff
inet 192.168.1.12/24 brd 192.168.1.255 scope global dynamic noprefixroute eth0
valid_lft 604765sec preferred_lft 604765sec
inet6 fe80::534b:4e:7aca:ece1/64 scope link noprefixroute
valid_lft forever preferred_lft forever

Hi,

Ok. Please check comment #12.

This issue is kind of new. We didn’t receive such error symptom before so need your help to add debug print to driver.

I’m busy trying to build the source. The instructions at https://docs.nvidia.com/jetson/l4t/index.html#page/Tegra%20Linux%20Driver%20Package%20Development%20Guide/kernel_custom.html isn’t very clear.

I’m compiling the kernel on the Jetson Nano. I have compiled the kernel itself in the kernel-4.9 folder. But that did not compile the modules in the nvidia folder.

Also, the instructions at the above linked page then tells you how to copy the kernel when you are compiling in a host PC and want to integrate that new kernel into sdkmanager. It does not tell you how to install the kernel when you compiled on the Nano.

My suggestion is download the kernel source from here to your host.

and follow the instructions on the document to cross compile the kernel on your host.

I have figured out how to compile on the Nano. Much easier to do it this way when ssh-ing into the Nano.

This is what I have figured out so far:
In the function rtl8168_get_mac_address I have found that:
tp->mcfg = 29

And the code:

            *(u32*)&mac_addr[0] = rtl8168_eri_read(ioaddr, 0xE0, 4, ERIAR_ExGMAC);
            *(u16*)&mac_addr[4] = rtl8168_eri_read(ioaddr, 0xE4, 2, ERIAR_ExGMAC);

Does return the 00:00:00:00:00:01 mac address.

I have not yet figured out what is going wrong in rtl8168_eri_read/rtl8168_eri_read_with_oob_base_address

As a side note, if I understand the data being printed then:

(u32)&mac_addr[0] = rtl8168_eri_read(ioaddr, 0xE0, 4, ERIAR_ExGMAC);

Returns “0”

(u16)&mac_addr[4] = rtl8168_eri_read(ioaddr, 0xE4, 2, ERIAR_ExGMAC);

Returns “256”

I don’t know if this helps, but this is the contents of /proc/net/r8168/eth0/driver_var

Dump Driver Variable
Variable Value


MODULENAME r8168
driver version 8.045.08-NAPI
chipset 29
chipset_name RTL8168H/8111H
mtu 1500
NUM_RX_DESC 0x400
cur_rx 0x2372
dirty_rx 0x2372
NUM_TX_DESC 0x400
cur_tx 0x13a6
dirty_tx 0x13a6
rx_buf_sz 0x5f3
esd_flag 0x0
pci_cfg_is_read 0x1
rtl8168_rx_config 0xcf00
cp_cmd 0x20e1
intr_mask 0x115
timer_intr_mask 0x4000
wol_enabled 0x1
wol_opts 0x20
efuse_ver 0x3
eeprom_type 0x0
autoneg 0x1
duplex 0x1
speed 1000
eeprom_len 0x0
cur_page 0x0
bios_setting 0x0
features 0x2
org_pci_offset_99 0x4
org_pci_offset_180 0x10
issue_offset_99_event 0x0
org_pci_offset_80 0x42
org_pci_offset_81 0x0
use_timer_interrrupt 0x1
HwIcVerUnknown 0x0
NotWrRamCodeToMicroP 0x0
NotWrMcuPatchCode 0x0
HwHasWrRamCodeToMicroP 0x1
sw_ram_code_ver 0x18
hw_ram_code_ver 0x18
rtk_enable_diag 0x0
ShortPacketSwChecksum 0x0
UseSwPaddingShortPkt 0x0
RequireAdcBiasPatch 0x1
AdcBiasPatchIoffset 0x56a8
RequireAdjustUpsTxLinkPulseTiming 0x1
SwrCnt1msIni 0x7bd
HwSuppNowIsOobVer 0x1
HwFiberModeVer 0x0
RequiredSecLanDonglePatch 0x0
HwSuppDashVer 0x0
DASH 0x0
dash_printer_enabled 0x0
HwSuppKCPOffloadVer 0x0
speed_mode 0x3e8
duplex_mode 0x1
autoneg_mode 0x1
aspm 0x1
s5wol 0x0
s5_keep_curr_mac 0x0
eee_enable 0x0
hwoptimize 0x0
proc_init_num 0x1
s0_magic_packet 0x0
HwSuppMagicPktVer 0x2
HwSuppCheckPhyDisableModeVer 0x2
HwPkgDet 0x0
random_mac 0x0
org_mac_addr 00:00:00:00:00:01
perm_addr 00:00:00:00:00:01
dev_addr 00:00:00:00:00:01

This reminds me an old issue here. Could you check if that helps?

I saw that firmware update, but as my board has a 2919 date code, and the link/activity LEDs work, I deduced that my board was not affected by that bug.

I’ll try it anyway and see if it does anything.

I have found that in rtl_eeprom.c(not at pc so that file name might be mis spelled), the magic number check fails when testing the eeprom. The value read is 0x0000 instead of the expected value. I assume that it fails to read, returns zero and the driver assumes “no eeprom”.

I will try to follow the trail back further, but the eeprom read calls appear to be some kind of system hooks or something strange, if I recall correctly.

Somehow your device does not read the mac from eeprom.

From our nano, we go through below path.

[ 2.379393] r8168 0000:01:00.0: enabling device (0000 → 0003)
[ 2.379409] r8168 Gigabit Ethernet driver 8.045.08-NAPI loaded
[ 2.393522] r8168 0000:01:00.0 (unnamed net_device) (uninitialized): Invalid ethernet address 00:00:00:00:00:00, trying device tree node
[ 2.393608] r8168 0000:01:00.0 (unnamed net_device) (uninitialized): Found valid ethernet address 00:04:4b:e5:5b:93 from device tree

I have ran the patch. I did patch, but still same after reboot.

If I do a “dtc -I fs /sys/firmware/devicetree/base | grep mac” then I do see the mac address in the device tree, so the driver isn’t pulling it from the device tree as it should?

  nvidia,ethernet-mac = "00:04:4b:e7:17:cc";
  		mac-address = [00 04 4b e7 17 cc];

If I do a “dtc -I fs /sys/firmware/devicetree/base | grep mac” then I do see the mac address in the device tree, so the driver isn’t pulling it from the device tree as it should?

Yes, I would suggest you can RMA your Nano to get a new one.

I think RMA is overkill.

I have narrowed the issue down more.

For some reason your board get a mac of 00:00:00:00:00 from trying to get it from eeprom. Then some code in the driver sees that that is an invalid mac. Then it pulls the mac from the device tree.

My board gets a mac of 00:00:00:00:01, which is a “valid” mac, so the mac from the device tree is not used. I have disabled the “valid mac check” and now get this"

[ 2.861051] r8168 0000:01:00.0 (unnamed net_device) (uninitialized): Invalid ethernet address 00:00:00:00:00:01, trying device tree node
[ 2.861130] r8168 0000:01:00.0 (unnamed net_device) (uninitialized): Found valid ethernet address 00:04:4b:e7:17:cc from device tree

And now my mac address is correct:

3: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP group default qlen 1000
link/ether 00:04:4b:e7:17:cc brd ff:ff:ff:ff:ff:ff

I’m starting to think that a broken 00:00:00:00:00:01 mac has been written to the OTP in the r8168 and that is being preferred?

I’m starting to think that a broken 00:00:00:00:00:01 mac has been written to the OTP in the r8168 and that is being preferred?

Yes, it is possbile.

I think RMA is overkill.

The decision is up to you. I just feel it is not good to use a module with wrong OTP fused.

After a hard power off and power on, instead of just an sudo reboot, the mac is 00;00;00;00;00;00 as it should be, so maybe the patch just required a full power down?

[ 2.866403] r8168 0000:01:00.0 (unnamed net_device) (uninitialized): Invalid ethernet address 00:00:00:00:00:00, trying device tree node
[ 2.866541] r8168 0000:01:00.0 (unnamed net_device) (uninitialized): Found valid ethernet address 00:04:4b:e7:17:cc from device tree

@marquitos Maybe also try the patch?

Thank you WayneWWW!

1 Like

Confirmed on my side. This script requires a cold boot instead of warm boot (commands).

2 Likes

is this use closed I have 4 jetson Nanos giving me the same problem

Hi quintinsmith334,

Please help to open a new topic with more details. Thanks