Jetson can not work with PCIe bus....

We try to connect the Jetson (TX1 and TX2) with a PCIe extender bus to our PCIe device and found out it seems not to understand the TS1 (ordersets) it gets.
We connected LeCroy and the same behavior had happened… (we see other side move to TS2 and Jetson remained in TS1 until it asserts PERST# again)

We could work with the Jetson only when there was an EP or a PCIe SWITCH connected directly to the PCIe socket.

My question - why is it? Do we do something wrong?
missing some kind of power?
Is it a known issue with the Jetson? signal integrity?
Did anyone tested Jetson with a more remote PCIe devices, not connected directly to the Jetson? with LeCroy?


Could you also share the kernel log and result of “lspci -vvv” with us?

Well, unfortunately, due to the fact the PCIe did not finish the Jetson is keeping reseting itself and never “wakes up” correctly to see anything…
All my visibility was via LeCroy to see what is going on on the PCIe bus

What exactly is the “PCIe extender bus” that you are referring to here? Can you please give a link where you have purchased that cable from?

Sorry - I refer to a simple PCIe bus that extend the connection instead of simply plug the device into the PCIe connector.
The fact is that LeCroy analyzer is connected directly to the PCIe connector and had the same problem :-(
We work with other RC with that PCIe bus and it works fine.

Is that extender cable spec compliant? It could probably to do with bad electricals. Possible to give the link where you purchased it from?

The cable works fine with all other setups (and there are many)
but in any case - lets talk about
Jetson -> LeCroy in the slot -> SSD
That setup had the same problem…
repeat PERST# due to Jetson remains in TS1 stage while the SSD moved to TS2 stage…

I think unless we have that cable, we can’t really say what the issue is. For now, it does look like it is something to do with bad electricals.

Is the Jetson more sensitive than other RCs?

Hi dalit.sagi,

Have you clarified the cause and resolved the problem?
Is this still an issue on your project?


It is still very unstable. with a re-timing device we manage to work, but it is unfortunately
We do not need it with other RCs :-(
Any idea how to do it otherwise?