Sounds like there is a way to return the value of the calculated checksum to the host CPU. Since the csum_data field is not used in the SCTP implementation, I assume there is not.

Also, I don't know how a network interface can indicate failure of checksum validation. It just drops the packet, right?
(This applies to the TCP/UDP paragraph starting at line 1194 as well)

How about this:

The driver of a network interface can indicate success of SCTP checksum validation by marking CSUM_DATA_VALID in csum_flags.

csum_flags

Sounds like there is a way to return the value of the calculated checksum to the host CPU. Since the csum_data field is not used in the SCTP implementation, I assume there is not.

I would say, currently there is no interface that provides the correct checksum. But that could change in the future. Also, this is the same text as for TCP and UDP.

Also, I don't know how a network interface can indicate failure of checksum validation. It just drops the packet, right?

No. My understanding is that the success of the validation is indicated by adding the CSUM_DATA_VALID flag. The failure of the validation is indicated by not adding the CSUM_DATA_VALID. However, the failure of the validation does not mean that the packet has an invalid checksum. Only that the hardware validation failed. That is why then the transport stacks perform the validation in software.

(This applies to the TCP/UDP paragraph starting at line 1194 as well)

How about this:

The driver of a network interface can indicate success of SCTP checksum validation by marking CSUM_DATA_VALID in csum_flags.

csum_flags

Thank for catching.

Sounds like there is a way to return the value of the calculated checksum to the host CPU. Since the csum_data field is not used in the SCTP implementation, I assume there is not.

I would say, currently there is no interface that provides the correct checksum. But that could change in the future. Also, this is the same text as for TCP and UDP.

Also, I don't know how a network interface can indicate failure of checksum validation. It just drops the packet, right?

No. My understanding is that the success of the validation is indicated by adding the CSUM_DATA_VALID flag. The failure of the validation is indicated by not adding the CSUM_DATA_VALID. However, the failure of the validation does not mean that the packet has an invalid checksum. Only that the hardware validation failed. That is why then the transport stacks perform the validation in software.

(This applies to the TCP/UDP paragraph starting at line 1194 as well)

How about this:

The driver of a network interface can indicate success of SCTP checksum validation by marking CSUM_DATA_VALID in csum_flags.

What about the updated wording. I now only add test regarding differences from TCP or UDP.

csum_flags

Thank for catching.

Fixed.

Add comma before or?

I don't know if there is a rule, but I saw sometimes a comma before an "and" or "or" in a list.

I find it harder to understand with this separation in two sentences. How about:

...the exact value of the checksum to the host CPU, its driver can mark CSUM_DATA_VALID in csum_flags as well as, for TCP and UDP, CSUM_PSEUDO_HDR and set csum_data to 0xFFFF hexadecimal to indicate a valid checksum.

Should "Internet" be replaced by "TCP/UDP"?

I find "in this case" not helpful here. Can it be omitted?

Add comma before or?

I don't know if there is a rule, but I saw sometimes a comma before an "and" or "or" in a list.

Fixed. It is called a serial comma.

I find it harder to understand with this separation in two sentences. How about:

...the exact value of the checksum to the host CPU, its driver can mark CSUM_DATA_VALID in csum_flags as well as, for TCP and UDP, CSUM_PSEUDO_HDR and set csum_data to 0xFFFF hexadecimal to indicate a valid checksum.

Good suggestion. Taken.

Should "Internet" be replaced by "TCP/UDP"?

It is called the Internet checksum, see RFC 1071.

I find "in this case" not helpful here. Can it be omitted?

Yes. Done.