A 10 Gb/s MAC sends back-to-back frames of mixed sizes. Now and then the PHY drops one, and the monitor sees Start on lane 2. Find the first bad column and the cause.
Hardware context: An Ethernet MAC and physical coding sublayer: frame builder and checker, reconciliation sublayer and block encoder. Build, check and pace Ethernet frames, and code them for the physical layer as IEEE 802.3 specifies.
The RS below lays packets onto the XGMII. It keeps Start on lane 0 with the deficit idle count: it deletes or inserts idle in the gap.
The MAC sends seven frames back to back: 64, 65, 66, 67, 65, 65 and 64 octets. The log says: ERROR: Start on lane 2.
Step 1: find the first column where a signal differs from a correct RS, and name it. Step 2: choose the cause from five bugs. Each is real and fails at a different column.
Step 3 is optional: fix the RS and run the tests.
Signals in the Timeline
| Signal | Width | Role |
|---|---|---|
test_number | 8 | input of each check |
Full specification
bit [7:0] fr [0:2047]; // the next frame
function automatic void rs_reset();
function automatic void rs_send(input int n); // queue one packet with an n-octet frame
function automatic [35:0] xg_col(); // {c[3:0], d[31:0]}, lane 0 in c[0] and d[7:0]
| Rule | Correct RS |
|---|---|
| Packet | Start (FBh, control), six 55h, D5h, the frame, Terminate (FDh, control) |
| Gap | the Terminate and 11 idles (07h, control): 12 characters, as the MAC asks |
| Start | on lane 0 only. If it would land on lane m, delete m idles from the gap, or insert 4 - m |
| Deficit idle count | deleting adds to it, inserting subtracts; delete only if the count stays at 3 or below |
| Columns | cycle N is column N; the signals are taken as the column goes out |
The count starts at 0. Both choices keep the long-run gap at 12 characters on average.
The 25GMII has the same logical function as the XGMII, with a faster clock (106.1). So one test sends a 25 Gb/s stream through the same RS, with the same rules.
Diagnose the Failing Run
The test below runs the buggy design. Read its log, then find where it first goes wrong in the waveform.
Log
Waveform
Click a signal at a cycle to select it. Or focus the waveform and use the arrow keys: Left and Right move one cycle, Up and Down one signal.
Step 3 is optional: fix the code in the editor, then run the tests.
Examples
Example 1
Four 64-octet Frames- Input
64 + 8 + 12 = 84 characters each
- Output
columns_differing_from_a_correct_rs = 0, frames_seen = 4, idle_characters = 1468, starts_off_lane_0 = 0
Every Start already lands on lane 0. Nothing is deleted or inserted.
Example 2
Eight 65-octet Frames- Input
each Start would land one lane late
- Output
columns_differing_from_a_correct_rs = 0, frames_seen = 8, idle_characters = 2768, starts_off_lane_0 = 0
The RS deletes one idle three times, then must insert three.
Example 3
Eight 67-octet Frames- Input
each Start would land three lanes late
- Output
columns_differing_from_a_correct_rs = 0, frames_seen = 8, idle_characters = 2752, starts_off_lane_0 = 0
Deleting 3 is allowed when the count is 0. Three single inserts bring it back to 0, and the pattern repeats.
Constraints and Contract
- Keep the four functions,
frand the queue. - Queue packets only through rs_send; the bench calls it when fewer than 16 characters wait.
- xg_col pads an empty queue with idle.
- Off limits: the names
mon,rng,tb,root, names that begin withref_and the modules of the bench.
Hints
Hint 1
Work the deficit idle count by hand for the first five frames. Which one has to insert idle?
Hint 2
If Start would land on lane 3, how many idles put it on lane 0 of the next column?
Hint 3
Watch dic at the column where the fifth packet (the 65-octet frame after the 67) is queued.
What the grader checks: The failing run is simulated live with the faulty RS. The first bad column and each option's symptom come from runs of each version on the same bench. The fix is checked on six fixed lists against an independent JavaScript model. One is a 25GMII stream, whose rules are the XGMII's (106.1). It is also checked on 400 random frames against a private reference RS, column by column.
8 checks on Icarus Verilog · server. Run the examples first, then check the full regression. Full checks use the server judge. Downloads contain your source only; samples require the simulation service. A passing result establishes the tested cases, not a proof for all possible inputs.
See all 1 steps and the final capstone