Skip to content
DenseChip
Learn
PracticeCode labInterview

DenseChip

Navigate

Menu

LearnPracticeCode labInterview

Tools

All toolsSV editorCode repoResume
DenseChip

Practice, courses, and chip-design tools in one focused workspace.

LearnPracticeCode labInterviewToolsLearning paths

© 2026 DenseChip

AboutContactPrivacyTerms
Skip to lesson

Course contents

  • How to Use This Course
  • Reading Paths
  • Concept Map
  • How to Read the Diagrams
  • Part 0: What Is a Protocol?10
  • Part 1: Timing and Clocking11
  • Part 2: Encoding and Framing12
  • Part 3: Handshakes13
  • Part 4: Flow Control17
  • Part 5: Transactions19
  • Part 6: Concurrency and Performance17
  • Part 7: Ordering13
  • Part 8: Sharing and Arbitration12
  • Part 9: Deadlock and Liveness13
  • Part 10: Reliability and Recovery20
  • Part 11: Bring-Up and Link Management10
    • Reset Semantics
    • Reset Sequencing Across Interfaces
    • Presence and Hot-Plug
    • Link Training
    • Capability Negotiation
    • Enumeration and Discovery
    • Link State Machines
    • Versioning and Compatibility
    • Protocols Compared: Bring-Up and Link Management
    • Drills: Bring-Up and Link Management
  • Part 12: Power8
  • Part 13: Shared State and Coherence9
  • Part 14: Security and Isolation7
  • Part 15: Specifying and Verifying11
  • Part 16: Protocols Dissected17
  • Part 17: Capstones5
  • Part 18: Interview Preparation6
Loading lesson

‹Part 11: Bring-Up and Link Management
  1. Topics
  2. First Principles of Hardware Protocols
  3. Part 11: Bring-Up and Link Management
  4. Drills: Bring-Up and Link Management

Part 11 · Bring-Up and Link Management · All levels

Drills: Bring-Up and Link Management

Interview questions, a whiteboard exercise and a ten-question checkpoint for Part 11.

On this page

    Answer each question out loud before opening it. If you miss more than three in the checkpoint, revisit the lessons linked at the end.

    Interview questions

    Check yourself What happens to an outstanding read when the target resets?

    It is lost; the requester waits for ever unless a timeout, a reset-aware wrapper, quiescing or a joint reset handles it.

    Check yourself Master out of reset before slave: what breaks?

    Requests into a target in reset are dropped and the master hangs; make start-up order-independent with an explicit readiness signal.

    Check yourself What must surprise removal do?

    Mark the device gone, complete every outstanding request with an error, free its resources, and let software recover.

    Check yourself Why does a fast serial link train?

    To find bits, symbols and lane order and tune equalisation to the channel before data can be decoded.

    Check yourself How do two partners settle on a speed?

    Each advertises its abilities and both take the highest common setting by the same rule; fall back if it fails to train.

    Check yourself How is a BAR sized?

    Write all 1s, read back; the bits that stay 0 give the window size.

    Check yourself Can a link carry traffic in recovery?

    No; packets wait, but state is kept and traffic resumes without loss.

    Check yourself What is the rule for reserved bits?

    Senders set them to 0; receivers ignore them.

    Whiteboard exercise

    Task. Specify start-up for a chip-to-chip link between two chips from different vendors, each with a different maximum speed and lane count, either of which may be reset or powered down independently. Cover reset behaviour, readiness, training, negotiation, the link state machine, and what happens to traffic when one side resets.

    Check yourself Open a model answer

    Reset: each side drives inactive outputs and sends nothing but training patterns until its link state machine reaches ACTIVE. States: DOWN, TRAINING (lock, deskew, equalise), INIT (credits and sequence numbers), ACTIVE, RECOVERY, LOW POWER, each with a timeout back to DOWN. Negotiation: training words carry a version, highest speed, lane count and feature bits; each side takes the minimum speed, minimum lanes and intersection of features, accepted only after several matching words; reserved bits set to 0 and ignored. Fallback: if the agreed setting fails to train, retry at the next lower speed or width. Order independence: no packets until credits are granted, so whichever side leaves reset first simply waits. One-sided reset: the partner sees training patterns or loss of lock, goes to DOWN, completes every outstanding request with an error, discards credits and replay state, and retrains; requesters also keep completion timeouts. Debug: expose the link state, time in each state and recovery counts.

    Checkpoint: ten questions

    Check yourself 1. What must a block drive while in reset?

    Inactive values: no valid signals, no requests.

    Check yourself 2. Name two reasons resets are released at different times.

    Separate per-domain synchronisation; waiting for PLL lock, calibration or firmware.

    Check yourself 3. Why does a credit-based link not care which side starts first?

    No credits, no sending: the sender waits for the receiver's initial credits.

    Check yourself 4. Orderly versus surprise removal?

    Orderly drains first; surprise leaves requests in flight that must be failed with errors.

    Check yourself 5. Why must bit lock come before symbol lock?

    Symbol boundaries can only be found in correctly recovered bits.

    Check yourself 6. Why is equalisation learned at training rather than configured?

    Each board's channel distorts differently and changes with temperature.

    Check yourself 7. A supports speed 3 with 2 lanes, B speed 1 with 16 lanes. Result?

    Speed 1, 2 lanes.

    Check yourself 8. How can two identical USB devices both start at address 0?

    The host enumerates one at a time, moving each to a unique address.

    Check yourself 9. Why should RECOVERY exit on the partner's pattern?

    So both ends leave together; local timers drift apart.

    Check yourself 10. Which side decides forward compatibility?

    The old side, by having ignored reserved and unknown content from version 1.

    Revisit

    • 11.1 Reset Semantics

    • 11.2 Reset Sequencing Across Interfaces

    • 11.3 Presence and Hot-Plug

    • 11.4 Link Training

    • 11.5 Capability Negotiation

    • 11.6 Enumeration and Discovery

    • 11.7 Link State Machines

    • 11.8 Versioning and Compatibility

    Keep going

    Next: Part 12: Power. How do we save power without losing data?

    Previous

    171 / 234

    Next: Part 12: Power

    Up next

    Up nextPart 12: Power

    Sign in to save progress

    Practice this lessonReport an error

    See also

    • Reset Semantics
    • Reset Sequencing Across Interfaces
    • Presence and Hot-Plug