Implement J1939 Soft Socket for SAE J1939 Transport Protocol in Python - #5011
Open
polybassa wants to merge 8 commits into
Open
Implement J1939 Soft Socket for SAE J1939 Transport Protocol in Python#5011polybassa wants to merge 8 commits into
polybassa wants to merge 8 commits into
Conversation
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #5011 +/- ##
==========================================
+ Coverage 80.05% 80.61% +0.56%
==========================================
Files 390 390
Lines 96810 97328 +518
==========================================
+ Hits 77499 78463 +964
+ Misses 19311 18865 -446
🚀 New features to boost your workflow:
|
Contributor
Author
|
@BenGardiner Please have a look. |
Contributor
|
Hi @polybassa , :) it looks like a much better implementation of what I tried out over in https://github.com/BenGardiner/scapy/tree/j1939_scanning . Thank you. I will try to rebase and port my scanner, DM , DM scanner and address claim implementation onto this branch and see how well those features still work. |
AI-Assisted: yes (GitHub Copilot)
AI-Assisted: yes (GitHub Copilot)
AI-Assisted: yes (Claude Sonnet 4.6)
AI-Assisted: no
…ostile transfers The soft socket handled the transfers its own tests produce, all of them under 100 bytes from a single well-behaved peer. Everything past that was either dropped or actively harmful, and none of it was visible because no test exceeded one CTS block, ran two senders, or sent a malformed frame. A payload of more than 1785 bytes needs more TP.DT packets than a sequence number can express, so building the announcement raised inside the scheduler thread after the TX state had already been set. The message vanished with a log line and the state machine stayed latched, which silently discarded every later send on that socket. send() now refuses such a payload, _J1939_TP_MAX_DATA finally being used for what it was defined for, and a failure anywhere in _begin_send resets the machine instead of wedging it. Reception was a single set of rx_* attributes, so a second announcement threw away the transfer in progress. Since a busy J1939 bus has several ECUs broadcasting at once, a monitor built on this socket lost most of what it saw. Sessions now live in a dict keyed by the (source address, destination) pair the protocol itself uses, capped so a hostile bus cannot grow it without bound, and a peer that asks for a second PGN while one is running is refused with an abort rather than displacing it. Frames from the bus are no longer taken at face value. An announcement whose size and packet count cannot describe a message is refused instead of delivering an empty payload; a CTS naming a packet outside the message is aborted instead of indexing the buffer backwards and emitting sequence number 0; and CTS, acknowledgement and abort frames must now name the PGN of the session they claim to be part of. tx_peer_sa is cleared when a session ends, so a node that took part in an earlier transfer can no longer abort an unrelated broadcast. The rest are smaller: the receiver honours the max_packets of a request and issues a CTS per block instead of authorising everything at once, close() derives its drain budget from what is left to send rather than truncating any broadcast longer than two seconds while __del__ no longer drains at all, basecls is used for delivered messages and recv_raw returns the payload, transport frames carry the caller's priority as single frames already did, a source address of 0xFF warns because it cannot legally appear on the wire, and a CAN socket that goes away closes the J1939 socket instead of leaving a caller blocked in recv() forever. The twelve new cases in the campaign each fail on the code before this commit and pass after it. The existing 185 are untouched and still pass. AI-Assisted: yes (Cursor) Co-authored-by: Cursor <cursoragent@cursor.com>
Four things the first pass got wrong or left rough. send() measured the payload differently from the code that transmits it, so the size guard and the wire could disagree for a message whose data was not bytes; both now go through one helper. The check for a CAN socket that has gone away was written twice in can_recv, once at each end. The new basecls parameter and the per-peer session model were undocumented. And the priority a caller asks for now reaches the TP.CM and TP.DT frames of a multi-packet message, which is a deliberate change of default from the 7 the code used to hardcode for TP.DT, so the docstring says so. AI-Assisted: yes (Cursor) Co-authored-by: Cursor <cursoragent@cursor.com>
polybassa
force-pushed
the
j1939_soft_sockets
branch
from
August 13, 2026 07:27
870efc6 to
ef10b34
Compare
… hid Four handlers cancelled a scheduler timeout inside try/except/pass, which Codacy flags and which hides a real failure as readily as the expected one. The expected one is narrow: TimeoutScheduler raises Scapy_Exception when a timeout has already fired or been cancelled, which races normally against the state machine dropping it. One _cancel helper now does that in the five places that needed it, logging anything else at debug level, and send() sets sent_time behind an isinstance check rather than catching the AttributeError a non-packet would raise. Building a connection abort was written out five times and refusing a session three times; both are helpers now, which is what made it obvious that the check for a TP.DT past the authorised block can never fire: the next CTS is sent from the same handler that completes a block, so the window it guards does not exist. Writing the test for it is what showed that, and both the branch and the test are gone. The four new cases cover what had no test: the session table filling up and answering with 'system resources', a stalled reception aborting with 'timeout' once its wall-clock ceiling passes, a sender aborting a reception it started, and close(timeout=0) as the way to give up on a transfer on purpose. AI-Assisted: yes (Cursor) Co-authored-by: Cursor <cursoragent@cursor.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
AI-Assisted: yes (GitHub Copilot)