The slave device cannot receive data from the master device

[Jan9401]

Use STM32F7 as the slave and Linux as the host. Linux supports hardware timestamps, and the host executes commands. The slave executes FlexPTP and does not receive any data packets sent by the host. I haven't had any ideas after debugging for a while, can you provide some ideas? about IGMP

[epagris]

Hi,

Have you given a try to the NUCLEO-F746ZG example? It should work right out-of-the-box, you can even download the precompiled binaries in case you don't want to build it from scratch.

If there's a router (or a capable managed switch) between the linuxptp machine and the flexPTP device, and UDP (L4) PTP operation is required, then the lack of IGMP processing (or messaging) might cause the flexPTP device to be locked off from receiving PTP-messages. If you're uncertain, you can always check IGMP-messages with Wireshark.

Do you use matching profiles at both the flexPTP and linuxptp? If unsure, try the default profile.
Do you see anything if you turn on timestamp, default and bmca logging? The first thing to kick in is the BMCA, since it determines the role of the device. You should see the state transitions with the logging turned on.
Are there any messages transmitted by the flexPTP device? You can easily check it with Wireshark, by connecting the flexPTP device directly to your computer.

Hope this helps! :)

[Jan9401]

Thanks for you response, I used NUCLEO-F467ZG example,only as slave, Do you mean two EVB boards, one for the host and one for the slave?

PC and EVB boards are on the same switch， I understand if the IGMP protocol is not necessary

I am using the default profile file in the LinuxPTP folder， bmca logging is disable, How to open BMCA log

[epagris]

Ok, then, are you using an STM32F4 or an STM32F7 device? (In your first message you wrote F7.) Although, you can start the F4 one on your F7 device for sure, it involves more likely less debugging to use the F7 example as a starting project, initially.

It's fully OK to test flexPTP against linuxptp and you don't necessarily need two flexPTP devices. (Sorry, I don't really know to which part of my previous response your question "Do you mean two EVB boards, one for the host and one for the slave?" is referring to.)

You can turn on BMCA logging with ptp log bmca on as described here.

Please try to check on the actual Ethernet communication with Wireshark.

[Jan9401]

Sorry, I made a mistake. It's not f476, it's F746. Ok I may have misunderstand. It's fully OK to test flexPTP against linuxptp. I have compiled the Linux PTP executable file and the network interface wlp0s20f3, so I can execute it directly /Flexptc_linux-i wlp0s20f3, but there is no response. What should be the correct approach

[epagris]

OK, so you're trying to run the flexPTP-linux example. Mind that it's an experimental one.

Just from the naming, wlp0s20f3 reads as a wireless (wlan) interface for me. Can you confirm, that it's a wired one?

[Jan9401]

I try flexptp-linux ,but i am failed. What are the correct steps

[epagris]

(First, check that your NIC supports PTP. Do you have the /dev/ptp<n> device files?)

1. Start flexptp_linux by specifying the right interface by passing its name, eg.: sudo flexptp_linux -i eth0. (Adjust executable path and network interface if needed.) At this point you should definitely see some log being printed.
2. Set the correct profile. If you want to use the default profile, let flexPTP load it by issuing the ptp profile preset default command. After setting the profile you should see a summary of the profile settings.

I highly recommend reading the docs.

[Jan9401]

I have /dev/ptp , This is the command I am executing, But the results in your document are different. Is there a problem with my configuration?

[epagris]

This is just fine, according to the log, flexPTP has switched into Master mode. You can check the actual messages with Wireshark.

However, you have not specified any profile setting, so I assume flexPTP has started up in the gPTP profile (due to this here). Does your other device also operate in the same, default profile? If unsure, issue a ptp profile preset gPTP command on both flexPTP instances.

These are logs from flexPTP instances (operating in gPTP profile) running on an MCU (left) and on Linux (right) with the MCU-based being the slave and the Linux-based the master:

[ym597061128]

I noticed there is only one ptp_config.cfg file in your flexlinux directory. Are you using this default cfg file? Also, could you capture and provide the commands you executed?

[Jan9401]

I used Wireshark to capture the package and found that only the host side sends delay requests

[epagris]

Hi @ym597061128,

ptp_config.cfg is just an example config file. The contents of any configuration file supplied to the flexPTP (on Linux) is just executed line-by-line as CLI commands. The config file is only taken in to account if specified with the -c parameter.

In the particular example shown above I did not pass any configuration to the flexPTP so it started up in gPTP mode (being the default startup mode for the flexPTP-linux example).

[Jan9401]

Why does the path_delay_dequest keep being sent as shown in the above picture?

[epagris]

@Jan9401:

Those ones should be Peer_Delay_Requests. It's normal operation that a gPTP instance probes the network for connected gPTP compatible peers. If a peer was present it would likely respond those messages ultimately triggering the full synchronization algorithm to kick in.

The same messages look like these below. Which version of Wireshark are you using? (You have some missing fields throughout the PTP dissection.)

[Jan9401]

There's something I don't quite understand in the program. The host needs (S.master. p2pSlave. state==PTP-P2PSS_SESTABLISHED) | | (! (S.profile. flags&PTP-PF.ISSUE_SYNC/FON_CompliantSLAVE-ONLY-INuP2P) to send sync_cessage in P2P mode. The ptpmaster_p2p_stlave_deported function will change the state variable to PTP-P2PSS_STABLISHED. But I didn't see where it was reported. This place is quite confusing