Skip to content

fix(tcp): MPTCP DSS picks 4- vs 8-octet Data ACK/DSN from the value instead of the a/m flags #1215

Closed

Description

@JarryShaw

Describe the bug
transport/tcp.py:2967,2970 set flag_a/flag_m from bit_length() > 32, and Data_MPTCPDSS does not store the a/m flags at all.

To Reproduce
An 8-octet Data ACK carrying a small value, 1e0c2003 0000000000000005, rebuilds as 1e082001 00000005, and the Data Offset drops from 8 to 7.

Expected behavior
The parsed a/m flags are kept and they decide the field width.

Found by the transport round-trip audit (#1202). The repro is from the audit report and I have not re-run it yet.

Activity

  1. added
    fixPull requests that fix a defect (fix: subject prefix)
    pendingAccepted and unstarted - nobody is on it and nothing blocks it
    on Oct 7, 2026
  2. JarryShaw commented on Oct 7, 2026

    @JarryShaw
    OwnerAuthor

    A worker is on it now.

  3. added
    wipWork in flight - a covering PR is open or an agent is actively on it
    and removed
    pendingAccepted and unstarted - nobody is on it and nothing blocks it
    on Oct 7, 2026
  4. removed
    wipWork in flight - a covering PR is open or an agent is actively on it
    on Oct 7, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    fixPull requests that fix a defect (fix: subject prefix)

    Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions