Skip to content

SINTRAN Timer Mechanism Analysis - Root Cause of HDLC Transmission Failures

Timer Configuration Discovery

HDLC Device Initialization (Lines 9678, 9725, 9772, etc.)

X21OP;POFTO;0;HDTM2;0;-2;IOX 1640;HDSTA;FALSE    % HDLC OUTPUT-DATAFIELD 1
X21OP;POFTO;0;HDTM2;0;-2;IOX 1660;HDSTA;FALSE    % HDLC OUTPUT-DATAFIELD 2
X21OP;POFTO;0;HDTM2;0;-2;IOX 1700;HDSTA;FALSE    % HDLC OUTPUT-DATAFIELD 3

Key Finding: HDLC timeout timer initialized to -2 (not -4)

Timer Value Interpretation

In SINTRAN III timer systems, negative values typically represent countdown timers in seconds: - -2 = 2-second timeout for HDLC - Timer decrements from -2 to 0, then triggers POFTO timeout interrupt

Interrupt Handler Analysis

HOINT - Main HDLC Interrupt Handler

104033   HOINT: 0=:TMR                    % RESET TIMER
104034          T:=HDEV+RTTS; *EXR ST    % READ STATUS  
104037          A=:HASTAT                % SAVE STATUS
104046          0=: ACTSW                % Clear active switch
104102          IF A/\ "SILFO+TXUND" = 0 THEN
104104                XRETRY=:RTDYN; A:=0; CALL SADTS

Critical Line 104033: 0=:TMR - Timer is RESET on every interrupt

POFTO - Timeout Handler

104117   POFTO: X:=OMSG; 0=:DCBX
104121          A:=CMODI; T:=HDEV+WTTC; *EXR ST
104125          A:=0; CALL SADTS; A:=ETOU1; GO FAR BACKX

Line 104125: A:=ETOU1 - Sets error code ETOU1 (timeout error)

Timer Mechanism Flow

Normal Operation

  1. Frame Sent → Timer starts counting down from -2 (2 seconds)
  2. Response Received → HOINT triggered → Timer reset via 0=:TMR
  3. Process Repeats for next frame

Timeout Scenario

  1. Frame Sent → Timer starts countdown (-2, -1, 0)
  2. No Response in 2 seconds → POFTO interrupt triggered
  3. Error Set → A:=ETOU1 (timeout error)
  4. Connection Terminated → GO FAR BACKX

Correlation with Observed Delays

Processing Delay Timeline

Time Period Frame Type Processing Delay Timer Status
Early Session 16-byte frames 600-1000ms ❌ EXCEEDS 2s TIMEOUT
Mid Session 40-byte frames 2000-3000ms ❌ EXCEEDS 2s TIMEOUT
Late Session 52-byte frames 7000-9000ms ❌ SEVERELY EXCEEDS 2s TIMEOUT

Critical Example - Frame Causing Timeout

SENT: [23:26:31.575] Machine 102 → 100 (52 bytes)
RECEIVED: [23:26:40.568] Machine 100 (DELAY: 8993ms = 8.99 seconds)

8.99 seconds >> 2.0 seconds = TIMEOUT TRIGGERED IMMEDIATELY

Root Cause Analysis

The Timeout Cascade

  1. Machine 100 Processing Degradation

    • Large frames (>40 bytes) cause progressive CPU bottleneck
    • Processing time grows from 600ms → 9000ms over session

  2. SINTRAN Timer Expiration

    • Fixed 2-second timeout cannot adapt to processing delays
    • Timer expires at 2000ms, but frame needs 8993ms to process

  3. Connection Termination

    • POFTO handler sets ETOU1 error code
    • Connection marked as failed and terminated
    • No recovery mechanism for timeout situations

Multi-Buffer Frame Impact

Large X.25 frames split across multiple buffers experience: - Assembly Overhead - Time to concatenate 2-3 buffer parts - Processing Complexity - Larger frames require more CPU cycles - Memory Allocation - Dynamic buffer management delays - Queue Congestion - Accumulated processing backlog

Solution Implications

Current Timer Configuration Issues

  1. Fixed Timeout - No adaptive timing based on frame size
  2. No Retry Logic - Single timeout terminates entire connection
  3. Processing Bottleneck - Machine 100 cannot handle large frame processing efficiently

Required Optimizations

  1. Increase Timer Value - Change from -2 to -10 or -15 seconds
  2. Fix Processing Bottleneck - Optimize Machine 100's frame processing
  3. Add Retry Logic - Multiple timeout attempts before connection termination

Conclusion

The 2-second SINTRAN timer is the definitive root cause of HDLC transmission failures.

Processing delays of 600ms-9000ms for ALL frames exceed the hardcoded 2-second timeout, triggering POFTO error handler and connection termination. Even the shortest processing delays (600ms) are well within the timeout range, explaining why the communication is so unreliable.

Critical Finding: With a 2-second timeout, virtually ALL observed frame processing times (600ms-9000ms) are at risk of timeout. The system was designed for much faster response times than what Machine 100 can deliver.

Fix Priority: Either optimize Machine 100's frame processing to under 2 seconds, or increase SINTRAN timer threshold to at least 10-15 seconds to accommodate realistic processing times.