Hi all,

this is a follow up issue to the discussion in #100. For context:

What I am still wondering, though, is why my setup works on the hardware ASIC without having to apply the instructions in the hw directory and without acquiring the bf-driver from the RDC. I documented the instructions for my setup here. I installed the open-p4studio version on a switch that previously had the Intel bf-sde installed, so maybe something got cached, and this is why it works? Are the hw instructions still required?

I set SDE=/home/ihlef/open-p4studio and SDE_INSTALL=/home/ihlef/open-p4studio/install.
The existing Intel bf-sde version is located in /opt/bf-sde-9.13.0, but should not be used.
I unloaded the modules bf_kdrv_mod and bf_fpga_mod in the Intel version folder, and loaded the ones from open-p4studio.

The program I am running on the ASIC is P4TG, a traffic generator application. I started the switch using the open-p4studio run_switchd.sh file.
To verify if it works correctly, I configured a traffic generation stream that is sent to an external server.
On the external server, I monitor the interface with wireshark, and see that the generated traffic arrives on the interface.
Further, the generated traffic is recirculated once in the switch before sending out, where some header manipulation is applied.
In the monitored traffic, those manipulated headers are visible, so I conclude that the pipeline actually processes the generated traffic.
P4TG also measures the generated traffic rate and the incoming traffic rate. Both match the configured values, so there is actually traffic on the wire.

I attached the bf_drivers.log files. I hope this is the correct one for bf_switchd. I see many errors there, but no pointer to the SDE path of the Intel version. The startup (where the SDE paths are referenced) is in bf_drivers.log.2.txt. The other log files are populated during traffic generation and are included for completeness.

Thanks in advance!
Fabian

bf_drivers.log.0.txt
bf_drivers.log.1.txt
bf_drivers.log.2.txt