theater/cabletest
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
2166a456a3b73418cce89995cba0a6a29491edca
cabletest/docs/state.md
T

44 lines
4.7 KiB
Markdown
Raw Normal View History

Add in-repo project docs: goals, state, hardware, module-PHY transports, module diagnostics, and measurement lessons
2026-08-10 10:34:26 -07:00
# Current state
X520 live as the product NIC: rate buckets re-keyed to read time (one shared clock, one commit per drained batch) with the backward smear as a settled window — each completed bucket enters once, donates excess to the nearest earlier deficits once, locks for display when nothing later can refill it (per-sample recompute double-counted excess and read >20G on a 20G wire); timestamp machinery deleted (ts.go, SO_TIMESTAMPING, rx_filter=ALL check) after the audit confirmed the buckets were its sole consumer; testDriver ixgbe; RollBall mailbox hardened against orphaned-command completions (never sample status before a poll gap, value from the same block read, devad/reg echo required — a killed session's 7.60 read answered a PHY ID as 0x0006); full hardware pass: 20.00G flat both directions, zero errors, ECD 42m, SNR +7.4dB
2026-08-17 09:35:58 -07:00
## Where the code is right now
Add in-repo project docs: goals, state, hardware, module-PHY transports, module diagnostics, and measurement lessons
2026-08-10 10:34:26 -07:00
Docs cleanup: stale-claim conflicts fixed (e810 stamps/probe.go, CX-5 candidate title, committed-path perf table now the X520 20G result with E810-era rows marked historical, ice-pair ring note), irdma/ice notes moved to e810, state.md halved (done-checklist, RX-steering and standing-goals duplication out; ECD suppression fact into modules/fs), changelog framing flattened to present tense
2026-08-17 12:32:36 -07:00
**The committed tree is the X520 product implementation**: AF_PACKET datapath with read-time rate buckets and the backward excess-fill smear (nics/x520/ — rate is a headline, never a verdict; `smearWindow` 4 buckets = 64 ms display latency, resize after measuring the stall distribution: open-questions), the mixed FS+Wiitek diagnostics with the BCM ECD length path, `testDriver` = ixgbe. No hardware-timestamp machinery anywhere. The CX-5 is retired (nics/connectx-5/, "Status") and the E810 bit-bang investigation with it (nics/e810/).
Add in-repo project docs: goals, state, hardware, module-PHY transports, module diagnostics, and measurement lessons
2026-08-10 10:34:26 -07:00
Restructure docs: per-device directories under modules/ and nics/ with co-located datasheets, state centralized in state.md, prose tightened into lists and tables
2026-08-12 19:23:02 -07:00
## Stashes
Nothing lives outside the repo: the bench ixgbe driver source moves in tracked at kernel/ixgbe-sff/ (build artifacts gitignored, load-ixgbe repointed, module build verified from the new home — the reg_ops/ts_bench bench extras are finally versioned), the ice sff_i2c patch is vendored at docs/nics/e810/ice-sff-i2c.patch with its out-of-tree build recipe (TRACE_INCLUDE_PATH fixup, version-matched Makefile), the e810 NVM flash logistics land in the e810 README (exact package name, rollback -m B49691D578EC into the inactive bank, post-flash inventory failing until power cycle is the expected tell), and every ~/work/phydiag-work pointer becomes an inline statement; phydiag-work, ice-sff, e810nvm and the wiitekphy CLI (a duplicate of documented transport+protocol) are deleted
2026-08-19 16:27:48 -07:00
None. Every parked experiment is distilled into docs and its code dropped: the AF_XDP conversion in xdp.md, the CX-5 Wiitek-pair implementation in nics/connectx-5/ (including the MCIA gateway constants), the phydiag campaign across nics/e810/, nics/x710/, modules/fs/ and modules/fibergaga/.
Restructure docs: per-device directories under modules/ and nics/ with co-located datasheets, state centralized in state.md, prose tightened into lists and tables
2026-08-12 19:23:02 -07:00
## The box
X520 live as the product NIC: rate buckets re-keyed to read time (one shared clock, one commit per drained batch) with the backward smear as a settled window — each completed bucket enters once, donates excess to the nearest earlier deficits once, locks for display when nothing later can refill it (per-sample recompute double-counted excess and read >20G on a 20G wire); timestamp machinery deleted (ts.go, SO_TIMESTAMPING, rx_filter=ALL check) after the audit confirmed the buckets were its sole consumer; testDriver ixgbe; RollBall mailbox hardened against orphaned-command completions (never sample status before a poll gap, value from the same block read, devad/reg echo required — a killed session's 7.60 read answered a PHY ID as 0x0006); full hardware pass: 20.00G flat both directions, zero errors, ECD 42m, SNR +7.4dB
2026-08-17 09:35:58 -07:00
- Single usable PCIe slot (Gen4 x8): holds the **X520-DA2** (dual SFP+). Verify 5 GT/s ×8 trained at bringup (`lspci -vv`). The ConnectX-5 is out of the box.
Restructure docs: per-device directories under modules/ and nics/ with co-located datasheets, state centralized in state.md, prose tightened into lists and tables
2026-08-12 19:23:02 -07:00
- X710 on a CPU x4 port (Gen3 x4, ~31.5 Gbps/dir — enough for 2×10G full duplex despite the driver's worst-case bandwidth warning).
- Many CPU cores; goroutine-heavy designs welcome.
X520 live as the product NIC: rate buckets re-keyed to read time (one shared clock, one commit per drained batch) with the backward smear as a settled window — each completed bucket enters once, donates excess to the nearest earlier deficits once, locks for display when nothing later can refill it (per-sample recompute double-counted excess and read >20G on a 20G wire); timestamp machinery deleted (ts.go, SO_TIMESTAMPING, rx_filter=ALL check) after the audit confirmed the buckets were its sole consumer; testDriver ixgbe; RollBall mailbox hardened against orphaned-command completions (never sample status before a poll gap, value from the same block read, devad/reg echo required — a killed session's 7.60 read answered a PHY ID as 0x0006); full hardware pass: 20.00G flat both directions, zero errors, ECD 42m, SNR +7.4dB
2026-08-17 09:35:58 -07:00
- mstflint (AUR — not in the Arch repos) provides `mstlink`/`mstreg`/`mstconfig` for the CX-5, if it ever returns.
Restructure docs: per-device directories under modules/ and nics/ with co-located datasheets, state centralized in state.md, prose tightened into lists and tables
2026-08-12 19:23:02 -07:00
| Interface | Device | Role (rules: hardware.md) |
|---|---|---|
X520 live as the product NIC: rate buckets re-keyed to read time (one shared clock, one commit per drained batch) with the backward smear as a settled window — each completed bucket enters once, donates excess to the nearest earlier deficits once, locks for display when nothing later can refill it (per-sample recompute double-counted excess and read >20G on a 20G wire); timestamp machinery deleted (ts.go, SO_TIMESTAMPING, rx_filter=ALL check) after the audit confirmed the buckets were its sole consumer; testDriver ixgbe; RollBall mailbox hardened against orphaned-command completions (never sample status before a poll gap, value from the same block read, devad/reg echo required — a killed session's 7.60 read answered a PHY ID as 0x0006); full hardware pass: 20.00G flat both directions, zero errors, ECD 42m, SNR +7.4dB
2026-08-17 09:35:58 -07:00
| `enp1s0f0` + second port | X520 (ixgbe) | Test pair — FS + Wiitek, dispatched by EEPROM PN so either port may hold either module. **Stock ixgbe refuses the FS's honest EEPROM and its port has no netdev until `./load-ixgbe`** (allow_unsupported_sfp=1 + the sff_i2c/ETQF patches); the tool needs exactly 2 ixgbe interfaces |
Noise ripped out after the bench verdict (tightly-wrapped unshielded cables move victim SNR only 0.2-0.3 dB, recorded in measurement.md dead ends): noise.go deleted with the noiser, phase clock and panel lane; configureNoise/checkPrivFlag gone so the X710 pair is untouched; console noise column, NOISE links rows and lane colors removed; driverPair rehomed to system.go, carrierUp to phy.go; history anchors to run start (frame-rate sampling and draw-clock roll kept — they fix display latency for every lane); goals/hardware/state/nics docs updated; verified on hardware (20.00G, zero errors, ECD 41m, SNR +7.6)
2026-08-17 21:37:18 -07:00
| `enp3s0f0np0` / `enp3s0f1np1` | X710 (i40e) | Unused by the tool (was the noise pair; alien crosstalk measured ineffective — measurement.md). Two Wiiteks still seated (has been `enp4s0f*` across reboots). The FS cannot live here (firmware rejects the honest EEPROM, no `allow_unsupported_sfp` in i40e) |
Restructure docs: per-device directories under modules/ and nics/ with co-located datasheets, state centralized in state.md, prose tightened into lists and tables
2026-08-12 19:23:02 -07:00
| `enp88s0` | igc | LAN uplink, default route; sibling `enp89s0` is dark |
## Hardware
| Item | Status | Notes |
|---|---|---|
X520 live as the product NIC: rate buckets re-keyed to read time (one shared clock, one commit per drained batch) with the backward smear as a settled window — each completed bucket enters once, donates excess to the nearest earlier deficits once, locks for display when nothing later can refill it (per-sample recompute double-counted excess and read >20G on a 20G wire); timestamp machinery deleted (ts.go, SO_TIMESTAMPING, rx_filter=ALL check) after the audit confirmed the buckets were its sole consumer; testDriver ixgbe; RollBall mailbox hardened against orphaned-command completions (never sample status before a poll gap, value from the same block read, devad/reg echo required — a killed session's 7.60 read answered a PHY ID as 0x0006); full hardware pass: 20.00G flat both directions, zero errors, ECD 42m, SNR +7.4dB
2026-08-17 09:35:58 -07:00
| X520-DA2 | **Installed** in the single PCIe slot — **the product NIC** (nics/x520/) | Unmediated bit-bang I2C (patched-ixgbe `sff_i2c`), <25 ms RollBall, FS links and the ECD length path is proven; the smoothed-bucket rate design removes the need for hardware RX stamps |
All three stashes distilled into docs and dropped: xdp.md carries the AF_XDP record (ixgbe's 3 KB single-buffer XDP cap excludes the locked jumbo mix; XSK need_wakeup TX as the spin-free flow control, zero-copy-or-die, the stray-RX fill-ring trap, BTF-resolved stamp kfunc, wire-weighted size schedule as an unadopted option); connectx-5 README gains the PCICONF/ICMD/MCIA gateway constants so a rebuild needs no mstflint re-mine; e810 keeps the read-side transport map (cage-cache, bus=port, bus-5 phantom, 16-bit-addr hazard, latency as wire-truth) and NVM rollback pointers; x710 the >=30 ms EEPROM write pacing; fibergaga the empirical vendor-space finds; fs the inert-I2C-surface map; state.md stash section now empty
2026-08-19 14:42:13 -07:00
| ConnectX-5 | Out of the box — retired | Ran the two-Wiitek pair at near line rate with dual-end SNR (MCIA transport, recorded with its gateway constants in nics/connectx-5/), but the firmware quantizes every module mailbox read to ~150 ms steps and the FS never links behind mlx5 — nics/connectx-5/, "Status" |
X520 live as the product NIC: rate buckets re-keyed to read time (one shared clock, one commit per drained batch) with the backward smear as a settled window — each completed bucket enters once, donates excess to the nearest earlier deficits once, locks for display when nothing later can refill it (per-sample recompute double-counted excess and read >20G on a 20G wire); timestamp machinery deleted (ts.go, SO_TIMESTAMPING, rx_filter=ALL check) after the audit confirmed the buckets were its sole consumer; testDriver ixgbe; RollBall mailbox hardened against orphaned-command completions (never sample status before a poll gap, value from the same block read, devad/reg echo required — a killed session's 7.60 read answered a PHY ID as 0x0006); full hardware pass: 20.00G flat both directions, zero errors, ECD 42m, SNR +7.4dB
2026-08-17 09:35:58 -07:00
| FS SFP-10G-T-100 ×2 | One in the X520 test pair, one on the shelf | BCM84891L, documented, robust; ixgbe-only (unlinkable behind mlx5 — host-interface deadlock, confirmed module-typed, nics/connectx-5/ — and rejected by the X710). The FS+Wiitek mixed pair (FS length via ECD + Wiitek SNR) is the product configuration |
Noise ripped out after the bench verdict (tightly-wrapped unshielded cables move victim SNR only 0.2-0.3 dB, recorded in measurement.md dead ends): noise.go deleted with the noiser, phase clock and panel lane; configureNoise/checkPrivFlag gone so the X710 pair is untouched; console noise column, NOISE links rows and lane colors removed; driverPair rehomed to system.go, carrierUp to phy.go; history anchors to run start (frame-rate sampling and draw-clock roll kept — they fix display latency for every lane); goals/hardware/state/nics docs updated; verified on hardware (20.00G, zero errors, ECD 41m, SNR +7.6)
2026-08-17 21:37:18 -07:00
| Replacement Wiiteks | One in the X520 test pair, two seated in the unused X710, rest on the shelf | Originals bricked by register exploration — modules/wiitek/ trap first. IEEE-standard registers only (modules/README.md safe set); VCT templates are single-shot candidates on a sacrificial unit only |
X520 live as the product NIC: rate buckets re-keyed to read time (one shared clock, one commit per drained batch) with the backward smear as a settled window — each completed bucket enters once, donates excess to the nearest earlier deficits once, locks for display when nothing later can refill it (per-sample recompute double-counted excess and read >20G on a 20G wire); timestamp machinery deleted (ts.go, SO_TIMESTAMPING, rx_filter=ALL check) after the audit confirmed the buckets were its sole consumer; testDriver ixgbe; RollBall mailbox hardened against orphaned-command completions (never sample status before a poll gap, value from the same block read, devad/reg echo required — a killed session's 7.60 read answered a PHY ID as 0x0006); full hardware pass: 20.00G flat both directions, zero errors, ECD 42m, SNR +7.4dB
2026-08-17 09:35:58 -07:00
| Fibergaga SFP-10G-T-30M | On the shelf | Aquantia, RollBall, the documented oracle; the alternate length path if the pair mix ever changes |
Restructure docs: per-device directories under modules/ and nics/ with co-located datasheets, state centralized in state.md, prose tightened into lists and tables
2026-08-12 19:23:02 -07:00
| 10Gtek | In hand | Claims SFP-10G-SR, still copper RJ45; filler, not in the test set |
| E810 | Out of the box | Patched ice + `sff_i2c` remains useful only if it returns for read-side work |
Nothing lives outside the repo: the bench ixgbe driver source moves in tracked at kernel/ixgbe-sff/ (build artifacts gitignored, load-ixgbe repointed, module build verified from the new home — the reg_ops/ts_bench bench extras are finally versioned), the ice sff_i2c patch is vendored at docs/nics/e810/ice-sff-i2c.patch with its out-of-tree build recipe (TRACE_INCLUDE_PATH fixup, version-matched Makefile), the e810 NVM flash logistics land in the e810 README (exact package name, rollback -m B49691D578EC into the inactive bank, post-flash inventory failing until power cycle is the expected tell), and every ~/work/phydiag-work pointer becomes an inline statement; phydiag-work, ice-sff, e810nvm and the wiitekphy CLI (a duplicate of documented transport+protocol) are deleted
2026-08-19 16:27:48 -07:00
## Bench driver
Note Wiitek ixgbe RollBall client and flag ixgbe RX-steering as open investigation
2026-08-10 11:29:05 -07:00
Nothing lives outside the repo: the bench ixgbe driver source moves in tracked at kernel/ixgbe-sff/ (build artifacts gitignored, load-ixgbe repointed, module build verified from the new home — the reg_ops/ts_bench bench extras are finally versioned), the ice sff_i2c patch is vendored at docs/nics/e810/ice-sff-i2c.patch with its out-of-tree build recipe (TRACE_INCLUDE_PATH fixup, version-matched Makefile), the e810 NVM flash logistics land in the e810 README (exact package name, rollback -m B49691D578EC into the inactive bank, post-flash inventory failing until power cycle is the expected tell), and every ~/work/phydiag-work pointer becomes an inline statement; phydiag-work, ice-sff, e810nvm and the wiitekphy CLI (a duplicate of documented transport+protocol) are deleted
2026-08-19 16:27:48 -07:00
- **Patched ixgbe, bench build** (`kernel/ixgbe-sff/`, full driver source tracked; build artifacts gitignored) — out-of-tree module for the dev host: the product patch plus the bench-only extras (`reg_ops` value readback, `ts_bench`); `./load-ixgbe` rebuilds + swaps the module with `allow_unsupported_sfp=1`. The product subset (raw I2C + `sff_i2c`, ETQF steering, FCoE slot guard) is versioned at `kernel/ixgbe-sff.patch` and applied by `./mkimage` to every appliance kernel build — apply fails loudly when stable moves under it; port the patch then. Details: nics/x520/. Nothing else lives outside the repo; the retired diagnostics campaigns are distilled in these docs and their code is deleted.
Add in-repo project docs: goals, state, hardware, module-PHY transports, module diagnostics, and measurement lessons
2026-08-10 10:34:26 -07:00
Docs cleanup: stale-claim conflicts fixed (e810 stamps/probe.go, CX-5 candidate title, committed-path perf table now the X520 20G result with E810-era rows marked historical, ice-pair ring note), irdma/ice notes moved to e810, state.md halved (done-checklist, RX-steering and standing-goals duplication out; ECD suppression fact into modules/fs), changelog framing flattened to present tense
2026-08-17 12:32:36 -07:00
## Every boot
X520 live as the product NIC: rate buckets re-keyed to read time (one shared clock, one commit per drained batch) with the backward smear as a settled window — each completed bucket enters once, donates excess to the nearest earlier deficits once, locks for display when nothing later can refill it (per-sample recompute double-counted excess and read >20G on a 20G wire); timestamp machinery deleted (ts.go, SO_TIMESTAMPING, rx_filter=ALL check) after the audit confirmed the buckets were its sole consumer; testDriver ixgbe; RollBall mailbox hardened against orphaned-command completions (never sample status before a poll gap, value from the same block read, devad/reg echo required — a killed session's 7.60 read answered a PHY ID as 0x0006); full hardware pass: 20.00G flat both directions, zero errors, ECD 42m, SNR +7.4dB
2026-08-17 09:35:58 -07:00
Bugs-evident cleanup: cable-measure failure is fatal with its reason (failed/fail pretend-path and dead relinked plumbing deleted), bcm.command refuses partial SETs (0 or 5 params, stale-DATA trap made structural), BOOT gains a patched-ixgbe check naming stock-driver boots, panel fault rows matched by label and console rows strict on column count, phy.go split into sff/bcm/rollball and main.go's direction machinery into direction.go (pure moves), input events decoded via encoding/binary, gofmt; verified on hardware (20.00G, zero errors, ECD 41m)
2026-08-17 18:56:55 -07:00
1. `./load-ixgbe` — until then the FS port has no netdev (stock driver, error -95) and `driverPair("ixgbe")` cannot find 2 interfaces. cabletest's BOOT check names the condition (`patched ixgbe` fails when any bound port lacks `sff_i2c`).
Docs cleanup: stale-claim conflicts fixed (e810 stamps/probe.go, CX-5 candidate title, committed-path perf table now the X520 20G result with E810-era rows marked historical, ice-pair ring note), irdma/ice notes moved to e810, state.md halved (done-checklist, RX-steering and standing-goals duplication out; ECD suppression fact into modules/fs), changelog framing flattened to present tense
2026-08-17 12:32:36 -07:00
2. Re-apply host tuning (hardware.md) before trusting results.
Reference in New Issue Copy Permalink