PCIe doesn't work ( Jetson AGX orin / custom b'd)

Autonomous Machines Jetson & Embedded Systems Jetson AGX Orin
,
1

Hi,

Custom board (I designed it myself)

PCIe reference clock & bus doesn’t work (M.2M)
Devices connected via PCIe are as follows.

  1. RTS5420 USB HUB (Block#0 Lane#2)–> Normal operation (USB 4 port normal)
  2. M.2M (Block#0 C#4 Lane#4~#7) → Not working
  3. I210 (intel) (Block#1 C#5 Lane#0~#3)) → Not working (Driver install need)
  4. AQR113C (Marvell: 10G/MGBE Block#2 Lane#4) → Flashing not yet

The Device Tree configuration has not been applied yet.

I have attached the booting log file below.
BootingLog.txt (89.5 KB)

also, I attach the lsmod.txt file below
lsmod.txt (2.7 KB)

Thank you

2

Does same card work fine on NV devkit?

3

Hi,

Thank you very much.
I will try this test again.

4

Hi,

I think M.2M should work. (Designed identically to NVIDIA Reference Schematic)
M.2M PCIe H/W connection (M.2E not used)
PCIe Usage: PCIe Block#0, Controller #4, Lane#4~#7

When I enter the “lspci” command in the terminal, there is no output.

M.2M Signal oscilloscope measurement level.
M.2M Pin 69: floating state (PCIe)
PERST# (PCIE4_RST_N): Low state
PEWAKE# (PCIE_WAKE_N): High state
SMB_DATA, SMB_CLK: Serial Resistor No STUFF
SUSCLK (32KHz): 32KHz input normal
CLKREQ# (PCIE4_CLKREQ_N): Low state

thank you

5

Hi,

I don’t know what does that mean “should work”. Do you have the real NVIDIA devkit to test the M.2 card instead of “you think”?

This is to cross check with NV devkit result.

6

Hi,

I don’t know what does that mean “should work”
==> The M.2M related H/W circuit is designed identically to the NVIDIA-released circuit. So, I think this function should basically work normally.

Do you have the real NVIDIA devkit to test the M.2 card instead of “you think”?
==> I have an NVIDIA devkit (Third-party product), but it is defective so I can’t test it with it.

Thank you

7

Hi,

==> The M.2M related H/W circuit is designed identically to the NVIDIA-released circuit. So, I think this function should basically work normally.

This really does not matter to me. I have heard lots of users talking about same comment as what you said but turned out they are not same as devkit…

Please find a valid devkit for cross check.

1 Like
8

Dear WayneWWW,

This issue has been resolved.
By replacing NVMe, Access became successful.
It is believed that the cause was a defective NVMe.
Coincidentally, there were 3 NVMes, but 2 out of 3 were defective.

Thanks for your help.