Determination and assigning of MSS in TCP

[kisai19980124]

I am new to iPerf3 and recently checking the speed of the mobile network between my MacBook and an iPerf3 server.
I am curious about how MSS is determined if -M argument is not assigned. Is the MSS taking from the system?

BTW I am always getting "iperf3: error - unable to set TCP/SCTP MSS: invalid argument" error on my MacBook when assigning the MSS using "-M 1448". Does someone know how it is happening and is there any way to solve it?

[davidBar-On]

I am curious about how MSS is determined if -M argument is not assigned. Is the MSS taking from the system?

The TCP stack calculates the default MSS size. See for example this IBM explanation.

I am always getting "iperf3: error - unable to set TCP/SCTP MSS: invalid argument" error on my MacBook when assigning the MSS using "-M 1448".

On my Linux -M 1448 works o.k. What is the reported default "Control connection MSS" MSS when you run the client with -V?

[kisai19980124]

Thank you for your replying.

For the first question, thank you for the reference. Is this a standard function included in the TCP/IP protocol?
Sorry for my poor network knowledge.

For the second question,

What is the reported default "Control connection MSS" MSS when you run the client with -V?

The default MSS is 1448 if I didn't add MSS argument.
TCP MSS: 1448 (default)

Maybe it's the same problem as #779 (comment) ?

iperf3: error - unable to set TCP/SCTP MSS: Invalid argument
I've reproduced this problem as well, although interestingly I see it under IPv4 but not IPv6 in my testing thus far.

Though I couldn't find further information about this issue.

[davidBar-On]

Is this a standard function included in the TCP/IP protocol?

TCP SYN/ACK 3-way handshake is a standard TCP functionality. The MSS calculation depends on the specific TCP stack implementation, although I assume that at least most of the implementations does something similar to the described calculation.

Maybe it's the same problem as #779 (comment) ?

There is a good chance that it is, and there is there is a bug in the MacOs implementation. Maybe someone else have further indite into this.