#include "FrameDecoder.h" #include namespace udpscope { /** Fallback cycle period before the first inter-packet gap is known. */ static constexpr double kDefaultDt = 1.0e-3; /** * How far a chained burst prediction may sit from where arrival time says it * should be before the chain is abandoned and time is re-anchored on arrival. * * This is a backstop, not the primary mechanism: the packet counter normally * accounts for lost datagrams exactly, so the prediction and arrival agree. * It catches what the counter cannot describe — a producer restart (the * counter returns to zero), a counter that never advances, and a declared * sampling rate that does not match the producer's real one. A kernel draining * a backlog of queued datagrams can legitimately put the prediction a couple of * hundred milliseconds from arrival, so the threshold sits well clear of that. * Same value and same reasoning as ClockOffset::kRecalibThresholdS. */ static constexpr double kBurstResyncThresholdS = 0.5; /** * Narrowest a burst may be drawn, as a fraction of its nominal width, while a * leading timeline is being pulled back. Only a floor: the squeeze is normally * proportional to the excess and removes it in a single burst. See the sole use * site. */ static constexpr double kMinBleedFactor = 0.05; /** * Largest share of the wall time elapsed since a signal's previous burst that * that signal's next burst may advance its own timeline by, while a lead is * being pulled back. * * This, and not kMinBleedFactor, is what makes a lead converge. A fraction of * the NOMINAL burst width cannot: the floor still advances the timeline by * kMinBleedFactor * nominal per packet while the wall advances one packet * interval, so it diverges outright whenever nominal exceeds * (1 / kMinBleedFactor) packet intervals — a declared SamplingRate of 30 against * a producer really flushing 10 samples at 1 kHz put the trace 667 s ahead after * 1000 s of stream. Measuring the allowance against elapsed WALL time instead * bounds the advance below the wall's own advance for any declared rate, so the * lead strictly falls whatever the config says. Any fraction under 1 converges; * a half both converges quickly and leaves the burst visibly compressed rather * than frozen. */ static constexpr double kWallBleedFraction = 0.5; /** * A backward jump in producer hrt larger than this is a producer RESTART; a * smaller one is a reordered datagram. * * The two need opposite handling — a reorder must leave the hrt reference * untouched (its interval was already counted by the packet that overtook it), * a restart must rebase onto the new epoch or the signal never advances again — * and nothing but the size of the jump distinguishes them. A second of producer * time is orders of magnitude more than any reordering window a UDP path can * produce (a few packet intervals) and orders of magnitude less than a restart, * which drops hrt from the producer's whole uptime back to near zero. * * Deliberately NOT kBurstResyncThresholdS: that one asks how far a WALL-clock * prediction may sit from arrival, a different quantity in a different clock * that happens to be tuned for delivery jitter. Sharing the number would couple * two unrelated tunings. */ static constexpr double kProducerRestartS = 1.0; /** * The declared sampling rate, or 0 when there is none to trust. * * samplingRate arrives unvalidated from a signal descriptor on the wire. A * malformed +inf reaches the reciprocal as dt == 0, which makes a burst's * nominal width zero and the proportional squeeze compute 0.0/0.0 — and a NaN * factor is not caught by the floor, since every comparison against NaN is * false, so the whole burst is emitted as NaN. Rejecting it at the boundary * costs one test and removes the entire class. */ static double DeclaredRate(double samplingRate) { return (std::isfinite(samplingRate) && samplingRate > 0.0) ? samplingRate : 0.0; } void FrameDecoder::setSignals(const std::vector& signals) { signals_ = signals; state_.assign(signals_.size(), SigState{}); hrtFit_.reset(); } void FrameDecoder::reset() { state_.assign(signals_.size(), SigState{}); hrtFit_.reset(); } void FrameDecoder::beginFrame(const FrameView& f) { if (f.hrt != 0u) { hrtFit_.add(f.hrt, f.recvTime); } } bool FrameDecoder::packetBurst(uint32_t idx, uint32_t nElems, double wallNow, std::vector& tsOut) { SigState& st = state_[idx]; if (!st.lastPacketValid || wallNow <= st.lastPacketWall) { /* No previous arrival to span from, or time went backwards. Remember * this one and drop the samples rather than store them at made-up * spacing. */ st.lastPacketWall = wallNow; st.lastPacketValid = true; return false; } const double dt = (wallNow - st.lastPacketWall) / static_cast(nElems); tsOut.resize(nElems); for (uint32_t e = 0; e < nElems; e++) { tsOut[e] = st.lastPacketWall + static_cast(e + 1u) * dt; } st.lastPacketWall = wallNow; return true; } bool FrameDecoder::timestamps(const FrameView& f, uint32_t idx, std::vector& tsOut) { tsOut.clear(); if (idx >= signals_.size() || idx >= f.numSignals || f.counts == nullptr || f.values == nullptr) { return false; } const SignalMeta& d = signals_[idx]; const uint32_t nElems = f.counts[idx]; if (nElems == 0u) { return false; } const double wallNow = f.recvTime; SigState& st = state_[idx]; /* A repeated counter is a duplicated datagram — the same update arriving * twice because the host joined the multicast group on two interfaces, say. * The C client only de-duplicates fragments, so an unfragmented update * reaches us intact both times; emitting it again would double the values * and advance the timeline by a burst that never existed. Counter zero is * excluded because a producer that never sets one leaves it there. */ if (st.lastEmittedValid && f.counter != 0u && f.counter == st.lastCounter) { return false; } /* hasTimeSignal() bounds the index against the FRAME's signal count, but * the time signal's type code is read from our own table, whose size is * independent — a frame carrying more signals than the installed table * (briefly possible after a CONFIG change) would otherwise read past it. */ const bool hasTimeSig = d.hasTimeSignal(f.numSignals) && d.timeSignalIdx < signals_.size(); const uint32_t tIdx = hasTimeSig ? d.timeSignalIdx : 0u; const double tScale = hasTimeSig ? TimeSignalScale(signals_[tIdx].typeCode) : 1.0e-6; /* Rule 1: one stamp per element, straight from the time signal. */ if (d.timeMode == kTimeFullArray && hasTimeSig && f.counts[tIdx] >= nElems && f.values[tIdx] != nullptr) { const double* tv = f.values[tIdx]; const double t0 = tv[0] * tScale; (void) st.offset.map(t0, wallNow); const double base = st.offset.offset(); tsOut.resize(nElems); for (uint32_t e = 0; e < nElems; e++) { tsOut[e] = base + tv[e] * tScale; } return true; } /* Rule 2: anchor from the time signal, spread by the sampling rate. */ if ((d.timeMode == kTimeFirstSample || d.timeMode == kTimeLastSample) && hasTimeSig && f.counts[tIdx] >= 1u && f.values[tIdx] != nullptr) { const double prodSec = f.values[tIdx][0] * tScale; const double anchor = st.offset.map(prodSec, wallNow); const double rate = DeclaredRate(d.samplingRate); double dt = (rate > 0.0) ? (1.0 / rate) : 0.0; /* No rate declared. UDPSourceSession.cpp:522 leaves dt at zero here, * which stacks every element of the array on one instant — harmless for * a host-local consumer that only stores them, but this scope's ring, * decimator and trigger all require a signal's stamps to increase, and a * plot of N points at one X is not a trace. * * The spread is recoverable without a rate: consecutive anchors come * from the time signal, so their difference is the burst's true duration * in producer seconds, measured on the producer's own clock rather than * on arrival — immune to the bursty delivery that corrupts everything * arrival-derived. Divide by the counter gap for the same reason rule 3 * does: a lost datagram widens the anchor difference without widening * the array. Until a second packet arrives there is nothing to measure * and the elements do stack; that is one packet, not the whole run. */ if (!(dt > 0.0) && nElems > 1u && st.prevAnchorValid && prodSec > st.prevAnchorProdSec) { const uint32_t gap = (f.counter != 0u && f.counter > st.lastCounter) ? (f.counter - st.lastCounter) : 1u; dt = (prodSec - st.prevAnchorProdSec) / (static_cast(nElems) * static_cast(gap)); } st.prevAnchorProdSec = prodSec; st.prevAnchorValid = true; st.lastCounter = f.counter; tsOut.resize(nElems); for (uint32_t e = 0; e < nElems; e++) { tsOut[e] = (d.timeMode == kTimeFirstSample) ? (anchor + static_cast(e) * dt) : (anchor - static_cast(nElems - 1u - e) * dt); } return true; } /* Rule 3: accumulated scalar, based on declared sampling rate or hrt. * * When samplingRate is declared the inter-element step is exact and we * anchor from the end of the previous burst rather than from arrival time * or hrt. This makes the output immune to arrival jitter: even when the * kernel delivers two packets microseconds apart each burst starts exactly * one sample period after the previous burst ended. * * When samplingRate is absent we must derive dt from the hrt gap, which * requires the HrtRateFit to be ready. Until then we fall back to * packetBurst (arrival-time spanning), which is accurate during the normal * pre-burst delivery phase that precedes the fit becoming ready. * * A signal that has already produced a burst stays on this rule even when a * later packet carries a single sample — Accumulate mode flushes on a timer, * so a short cycle legitimately yields one. Dropping such a packet to rule 5 * would date it from arrival while its neighbours are chained, and would * leave lastCounter behind so the next real burst read the skip as a lost * datagram and reinstated a hole that never existed. A signal that has never * burst is a genuine scalar and is left to rule 5. */ if (d.numElements() == 1u && (nElems > 1u || st.lastEmittedValid)) { /* A rate that is not a finite positive number is no rate at all; see * DeclaredRate(). Such a signal takes the hrt branch below. */ const double declared = DeclaredRate(d.samplingRate); const double dt = (declared > 0.0) ? (1.0 / declared) : 0.0; if (declared > 0.0) { /* Where arrival time says this burst begins: its last element was * acquired just before the packet landed. */ const double arrivalAnchor = wallNow - static_cast(nElems - 1u) * dt; /* Chaining onto the end of the previous burst is immune to arrival * jitter — a kernel draining several queued datagrams microseconds * apart still yields contiguous timestamps. What a bare chain gets * wrong is loss: it closes the hole a dropped datagram left, and * every later sample is then dated early for the rest of the run. * * The wire says exactly how much is missing. counter increments * once per update, so a gap of g means g-1 lost packets, each * carrying (as far as we can tell) as many samples as the last one * we saw. Reinstating that duration keeps the chain honest without * consulting arrival time at all. */ double base = arrivalAnchor; double step = dt; if (st.lastEmittedValid) { /* Unsigned subtraction wraps, so this stays right across the * counter's own 2^32 rollover. * * A producer restart or a reordered datagram makes the wrapped * gap enormous, and this deliberately does NOT special-case * that: an absurd gap yields an absurd prediction, which the * arrival backstop below then rejects on its own. * * Clamping the gap first is not a harmless earlier version of * the same decision — it reaches the OPPOSITE answer. A clamp * that treats gap > kMaxCounterGap as unknowable has to fall * back to lost == 0, so the prediction becomes * lastEmittedEnd + dt: one sample period after the last burst, * which is exactly the shape of a healthy chain and therefore * lands INSIDE the arrival backstop, is accepted, and silently * closes an outage of arbitrary length. Letting the absurd gap * through produces an absurd prediction that the backstop * catches, and the burst re-anchors on arrival — which is the * right answer, and what * AccumulatedScalarSurvivesAProducerRestart pins down. The * arithmetic cannot overflow: gap and prevAccCount are both * bounded by 2^32-1, so lost is at most ~1.8e19, finite, and * always rejected. */ const uint32_t gap = f.counter - st.lastCounter; const double lost = (gap > 1u) ? static_cast(gap - 1u) * static_cast(st.prevAccCount) : 0.0; const double predicted = st.lastEmittedEnd + dt * (1.0 + lost); /* Backstop for what the counter cannot express: a producer * restart, a counter stuck at zero, or a declared rate that is * simply wrong. Beyond this the chain is not recoverable and * arrival time is the better of two bad answers. */ if (std::fabs(predicted - arrivalAnchor) <= kBurstResyncThresholdS) { base = predicted; } if (base <= st.lastEmittedEnd) { /* Re-anchoring here would step backwards, and the ring, the * trigger and the exporter all require a signal's stamps to * increase. Rejecting the correction outright is not an * option either: `predicted` is never less than * lastEmittedEnd + dt, so rejection would make the backstop * one-directional and let a timeline that runs FAST — two * hosts' crystals differ by tens of ppm, so this is certain * on a long session, not hypothetical — drift ahead of the * wall clock without bound. * * So compress instead of stepping back: start immediately * after the previous burst and spread this one out to * arrival. A single packet is drawn narrower than its true * width, and in exchange the timeline is back in step. */ if (wallNow > st.lastEmittedEnd) { step = (wallNow - st.lastEmittedEnd) / static_cast(nElems); base = st.lastEmittedEnd + step; } else { /* The timeline has run PAST arrival: our last burst is * dated later than the moment this packet landed, so * there is no room to spread into and no burst can end * on arrival without starting before it. Squeeze this * one instead, by the excess and by the wall time that * has really elapsed since this signal's last burst. * * Both terms are needed, and only the second one * converges. Within this call the wall clock is frozen * at wallNow, so ANY positive step increases the lead * measured at this instant; the lead falls only because * the wall advances BETWEEN packets. A step expressed * purely as a fraction of the nominal burst width * therefore diverges as soon as the nominal width * outruns the packet interval — with the floor alone, a * declared 30 Hz against a producer really flushing 10 * samples at 1 kHz gained ~0.67 s of lead per second of * stream, without bound. Capping the burst's total * advance at kWallBleedFraction of the elapsed wall time * makes it advance strictly slower than the wall for any * declared rate, so the lead strictly falls. * * The proportional term still does the fine work: when * the excess is smaller than a burst it removes it in * one packet. kMinBleedFactor only keeps that term * positive when the excess exceeds a whole burst. * * Steady state is a sawtooth, not a fixed offset: the * squeeze pulls the lead down, ordinary chaining resumes * on the very next packet and pushes it back up until * the prediction misses arrival by more than * kBurstResyncThresholdS. So the lead cycles between a * fraction of a millisecond and roughly that threshold — * bounded, which is what matters, but not zero. */ const double nominal = static_cast(nElems) * dt; const double excess = st.lastEmittedEnd - wallNow; double factor = 1.0 - excess / nominal; if (factor < kMinBleedFactor) { factor = kMinBleedFactor; } double advance = nominal * factor; /* A non-positive elapsed means the wall has not moved * since this signal's previous burst. Skipping the cap * then is deliberate and, more to the point, makes no * difference: forcing the cap to zero instead sends step * through the floor below to dt * kMinBleedFactor, which * is the same number the proportional factor already * yields once the excess exceeds one burst. Both leave * the same-tick case diverging; only real elapsed wall * time can bleed lead off, and a recv_time from * CLOCK_REALTIME (udps_client.c:120) does not repeat. */ const double wallElapsed = wallNow - st.lastEmittedWall; if (wallElapsed > 0.0) { const double cap = kWallBleedFraction * wallElapsed; if (cap < advance) { advance = cap; } } step = advance / static_cast(nElems); /* Unreachable with a finite positive dt — kept because * downstream monotonicity must not depend on that * argument holding for every value off the wire. */ if (!(step > 0.0)) { step = dt * kMinBleedFactor; } base = st.lastEmittedEnd + step; } } } tsOut.resize(nElems); for (uint32_t e = 0; e < nElems; e++) { tsOut[e] = base + static_cast(e) * step; } st.lastEmittedEnd = tsOut[nElems - 1u]; st.lastEmittedWall = wallNow; st.lastCounter = f.counter; st.prevAccCount = nElems; st.lastEmittedValid = true; return true; } /* No declared rate: need hrt-derived dt. */ if (!hrtFit_.ready() || f.hrt == 0u) { const bool ok = packetBurst(idx, nElems, wallNow, tsOut); /* Carry the warm-up's state into the hrt branch, or the handover * from one to the other is a discontinuity in both directions. * * The producer-clock reference (lastAccHrt, prevAccCount) matters * most. Without it the first hrt packet has no previous tick to * subtract, falls back to kDefaultDt for its inter-element step and * latches ClockOffset against wallNow - (nElems-1)*kDefaultDt. * kDefaultDt is only right when the burst happens to run at 1 kHz; * at 100 samples per 10 ms packet it is ten times too wide and the * latch lands 89 ms in the past — permanently, since it is below * ClockOffset's recalibration threshold. Seeding here means the * first hrt packet measures a real tick delta and latches correctly. * * The emitted-timeline reference (lastEmitted*) then only has to * cover residual disagreement, but it is what keeps the handover * MONOTONIC: packetBurst ends its burst at wallNow while the hrt * branch ends at wallNow - (nElems-1)*hrtDt, and without a previous * end to clamp against the first hrt packet steps the signal * backwards by up to a whole burst width. */ if (f.hrt != 0u) { st.lastAccHrt = f.hrt; st.lastAccValid = true; } st.prevAccCount = nElems; if (!ok) { return false; } st.lastEmittedEnd = tsOut[nElems - 1u]; st.lastEmittedWall = wallNow; st.lastCounter = f.counter; st.lastEmittedValid = true; return true; } const double rate = hrtFit_.ticksPerSecond(); /* Integrate short tick DELTAS. Never convert an absolute tick count, and * never subtract two such conversions. * * hrt counts from the producer's boot, so it is already ~1e11 ticks when * the scope attaches, while the fit is re-estimated on every packet and * wobbles by a few parts in 1e4. Any absolute hrt/rate therefore carries * that relative wobble multiplied by the whole elapsed epoch — tens of * milliseconds, moving in either direction from one packet to the next. * As a burst's position that is not merely imprecise, it is * NON-MONOTONIC: on a 2 h stream with ordinary scheduling jitter a few * percent of samples land before their own predecessor. * * A delta spans one packet, so its share of the wobble is microseconds, * and summing deltas keeps it there. ClockOffset then latches the * arbitrary epoch that leaves behind, exactly as it would have latched * the producer's boot epoch. */ double elapsed = 0.0; /* Whether lastAccHrt should take this packet's value. Only a packet * that legitimately defines the new front of producer time may move it; * see the backward case below. */ bool takeHrt = true; if (st.lastAccValid) { if (f.hrt > st.lastAccHrt) { elapsed = static_cast(f.hrt - st.lastAccHrt) / rate; } else { /* hrt went backwards. Two entirely different events look like * this and only the SIZE of the jump separates them. * * A small one is a reordered datagram: the packet that overtook * it already counted the interval it covers, so it must * contribute nothing — and must also leave lastAccHrt alone. * Letting it write lastAccHrt anyway (which is what this code * used to do unconditionally) rolls the reference back one * interval, so the NEXT packet's delta spans two and fabricates * a whole extra packet of producer time. It never heals: * ClockOffset would correct it, but the monotonic clamp below * discards every backward correction. A hundred swaps on a * 25 ms stream left the trace 3.5 s ahead, permanently. The C * client does not reorder for us — udps_client.c only COUNTS * counter gaps — so this is reachable on any real network. * * A large one is a producer restart: hrt drops from the * producer's whole uptime to near zero. Here the unconditional * write was the right behaviour and must be kept, because * refusing to regress would leave every subsequent packet below * lastAccHrt forever, elapsed permanently zero and the signal * frozen. Rebase, and reset the offset so it re-latches against * the new epoch instead of being dragged there by recalibration. */ const double backward = static_cast(st.lastAccHrt - f.hrt) / rate; if (backward > kProducerRestartS) { st.offset.reset(); } else { takeHrt = false; } } } st.accProdSec += elapsed; /* The flushes carry contiguous RT cycles, so the tick gap divided by the * number of cycles it spans is exactly one cycle period. That count is * NOT prevAccCount: elapsed spans every packet since the last one we * saw, so a lost datagram makes the tick gap wider without making * prevAccCount larger. Dividing by prevAccCount alone therefore returns * a period scaled by the whole counter gap — 2x for one lost datagram, * 11x for ten — which draws the recovery burst that many times too wide * and, because the burst is anchored on its LAST element, ends it in the * FUTURE (measured: +22.5 ms for one loss, +225 ms for ten, at 10 * samples per 25 ms packet). At 1% loss that mis-spaced 2.7% of all * samples. The declared branch already reads the counter for exactly * this purpose (`lost`, above); the hrt branch must too. * * Only a FORWARD gap counts. A backward or repeated counter is the * reorder case handled above, where elapsed is zero anyway. */ const uint32_t accGap = (f.counter != 0u && st.lastEmittedValid && f.counter > st.lastCounter) ? (f.counter - st.lastCounter) : 1u; const double cycles = static_cast(st.prevAccCount) * static_cast(accGap); /* Falling back to kDefaultDt is a last resort, not a default: see * SigState::lastHrtDt. The fallback is reached on the first hrt packet * of a producer restart (elapsed is zero because hrt went backwards) and * on a reordered datagram, and in both cases the wrong burst width is * latched into ClockOffset permanently — measured 13.5 ms of standing * displacement at 10 samples per 25 ms packet, 89 ms at 100 per 10 ms, * both below kRecalibThresholdS and so never corrected. */ double hrtDt; if (elapsed > 0.0 && cycles > 0.0) { hrtDt = elapsed / cycles; st.lastHrtDt = hrtDt; } else if (st.lastHrtDt > 0.0) { hrtDt = st.lastHrtDt; } else { hrtDt = kDefaultDt; } /* Anchor the burst's LAST element on arrival, not its first. The * packet's hrt is the tick count of sample 0 (UDPSourceSession.cpp:574), * so stepping forward from it is right — but ClockOffset latches * offset = wall - producerSec on its first call, and passing the raw * arrival would put sample 0 at the instant the packet carrying the * whole burst LANDED, dating every sample in it late by a burst. The * declared-rate branch above already anchors on the burst end * (arrivalAnchor), and two accumulated scalars in one scope, one with a * declared rate and one without, would otherwise sit a burst apart on a * shared X axis — 9 ms for 10 samples at 1 kHz, plain to see at a 200 ms * window. Since map() latches once, this is a constant shift applied at * latch and recalibration only; it changes no spacing. */ double base = st.offset.map( st.accProdSec, wallNow - static_cast(nElems - 1u) * hrtDt); double step = hrtDt; /* ClockOffset recalibrates once true drift passes its threshold, and a * recalibration can land behind where this signal already is. * Downstream requires increasing stamps, so step forward minimally — * but a bare forward step is one-directional, exactly the defect the * declared branch's squeeze exists to avoid. A backward wall step (an * NTP correction, a suspend/resume) would otherwise leave this signal * permanently ahead of the wall clock, since the recalibrated base is * behind lastEmittedEnd on every later packet too and the clamp keeps * discarding it. So cap the burst's total advance against the wall time * elapsed since this signal's previous burst, for the reason spelled out * at kWallBleedFraction: only that makes the lead bleed off. */ if (st.lastEmittedValid && base <= st.lastEmittedEnd) { const double wallElapsed = wallNow - st.lastEmittedWall; if (wallElapsed > 0.0) { const double cap = kWallBleedFraction * wallElapsed / static_cast(nElems); if (cap < step) { step = cap; } } base = st.lastEmittedEnd + step; } tsOut.resize(nElems); for (uint32_t e = 0; e < nElems; e++) { tsOut[e] = base + static_cast(e) * step; } if (takeHrt) { st.lastAccHrt = f.hrt; } st.lastAccValid = true; st.prevAccCount = nElems; st.lastEmittedEnd = tsOut[nElems - 1u]; st.lastEmittedWall = wallNow; /* Keep packetBurst's reference current even though this branch does not * use it. A single packet with hrt == 0 re-enters the warm-up branch * above, and packetBurst would otherwise span from whenever this signal * last took that branch — the whole session. Measured: after 153 hrt * packets, one zero-hrt packet emitted a burst starting 3.8 s in the * past, growing without bound with session length. */ st.lastPacketWall = wallNow; st.lastPacketValid = true; /* Same duplicate-datagram exposure as the declared branch: a host joined * on two interfaces receives every unfragmented update twice, and the * guard at the top of timestamps() can only fire if this branch leaves a * counter behind for it to compare against. */ st.lastCounter = f.counter; st.lastEmittedValid = true; return true; } /* Rule 4: PACKET burst with no time reference at all. */ if (nElems > 1u) { return packetBurst(idx, nElems, wallNow, tsOut); } /* Rule 5: plain scalar. */ tsOut.assign(1, wallNow); return true; } } /* namespace udpscope */