The audit's P1 robustness findings:
- _etu_candidate() validates a measured burst: requires >=2 usable
spacings, rejects a guard-dominated burst (all-zero data: 10-ETU LOW
runs + a 2-5 ETU HIGH guard) and a lone 10-ETU spacing. A failed
measurement leaves bit_samples unset instead of committing the
clock_skip*3 fallback.
- _confirm_etu() no longer counts a dropped frame (empty bits) as
parity-OK, and a rejected measurement clears bit_samples,
_etu_measured and _clock_skip_confident (was: kept sampling at the
rejected rate).
- Hunt ETU recovery keys off _etu_measured: handle_atr pre-seeds
bit_samples, so the old "bit_samples is None" condition never fired.
- resync_idle() waits on the DATA idle level (plus RST/VCC edges,
bounded) instead of spinning on wait({'skip': 0}) at one sample.
- ATR acceptance: a DATA-state 0x3B/0x3F is a candidate only before the
first APDU (_apdu_seen); a parsed ATR is ignored on any parity error
or when it is a bare TS+T0. A data run misread at a frame boundary
could fabricate an ATR and force clock_skip=372 for the capture.
- Inverse convention per ISO 7816-3 7.1/8.1: idle HIGH and a falling
start bit in both conventions, msb-first inverse data, raw moments
5-7 TS-pattern detection (a real inverse TS reads as 0x03 through
the direct reader), and a TS parity exemption (the inverse pattern
is odd over moments 2-10). Spec-verified; no inverse capture in the
corpus.
Validation: 117/117 unit tests (18 new); the 10-trace suite is
unchanged from v1.9.5 (APDU/CHKSUM/CONCAT/BAD_FCS identical,
RESULT: OK everywhere); compare_baseline.sh reports IDENTICAL.
VERSION -> 1.9.6.
- Decode the first post-header byte as SW1 before the ACK/data decision:
a case-2 rejection (6A82/6CXX) carries no ACK and used to be read as
command data, silently consuming the exchange and the next one.
- Short responses (Lr < Le): after the P3-byte read, split the buffer at
the first plausible SW1 followed by a plausible CLA, replay the tail and
flag BAD_FCS; previously the SW scan discarded the real SW, merged the
following exchanges and lost the exchange at EOF.
- P3=0 is Le=256 for an outgoing command: keep the response data (read
until SW) instead of discarding it, including data bytes equal to INS.
- ~INS ACK: one data byte per ACK via the new helpers; case-4 inline data
is appended (permissive; ISO 7816-3 §12.2 maps case 4 to GET RESPONSE).
- tests: four TestDecodeStep vectors (immediate SW, short read + split +
replay, Le=256 data incl. an INS-valued byte); 99/99.
- AGENTS v1.9.5 + T=0 section, README/README_RUS bullets; baseline
regenerated (xiaomi_mi_a1_coldboot 473 -> 475 APDUs, all else identical);
compare_baseline: IDENTICAL.
- tests: TestDecodeStep drives decode_step() with a scripted byte queue
(read/peek/consume stubbed over _replay + the queue; the re-align/PPS/
ATR/annotation sinks stubbed): 'CX'/'EX' exchanges frame and emit clean,
0x5F frames as a command (it was always eaten by the idle-noise gate
before v1.9.3), the 0x61 text-byte boundary is a one-byte suspicious-CLA
discard, the invalid-INS re-frame recovers a shifted SELECT, 0xFE still
takes the idle-noise path.
- tests: TestAtrHuntDetect drives handle_atr(): 0xC4 0xA4 / 0xC4 0x89
hand the header to DATA via _replay, CLA 0x00 stays excluded, an invalid
INS does not hand off.
- pd.py: the hunt command-detect guards peek_byte() returning None (a
capture ending right after a plausible CLA byte raised TypeError, seen
only by decode()'s crash protection);
test_eof_after_plausible_cla_is_clean pins the clean return.
- AGENTS v1.9.4 section; baseline regenerated (version + test listing;
compare_baseline: IDENTICAL); 95/95 unit tests.
- pd.py/AGENTS: the rejected '8X' range now reads 0x81-0x83/0x85-0x8F
(0x84 stays accepted); the 0x40-0x4F vs 0x50-0x5F citations are split
(chaining is ISO-only, TS 102 221 keeps b5=0); the bit-dense CLA list
gains 0xE7 (exact set verified).
- README_RUS.md: the invalid-INS note mirrored (the removed odd-LSB rule,
citation corrected to 5.1.2).
- tests: test_vs_reader_mirror (pd.py <-> tools/vs_reader.py drift guard),
test_odd_ins_set_pinned, the 0xE7 noise vector, and the 'EX' vector now
uses STATUS (e2 f2) as TS 102 221 Table 10.5 defines.
- tests/baseline.txt regenerated (test listing only); compare_baseline:
IDENTICAL.
- handle_pps(): flag the first post-PPS character; _read_byte_clk() re-measures
the CLK period there and accounts the measurement's 16 edges. The Alcor
reader upclocks 4.0 -> 4.8 MHz at PPS (FI/DI still 512/32), so the ATR
measurement was stale: 64 vs 53.33 samples/ETU, no post-PPS APDU framed.
- _wait_clk_rising(): project to the n-th edge (skip to target - half a
period, take one edge) instead of "skip n-2 cycles + 2 edges", which lost
~1 edge per call on fractional periods and slid the samples ~0.5 ETU early
by the parity bit. xiaomi 7698 -> 2, 4gmodem 7218 -> 0, S21p_sim2 394 -> 0
CHKSUM lines (previously misread as dirty-wire noise).
- CONCAT gate: EMBEDDED_CLA excludes the structural 0x4x/0x5x classes (payload
false hit on the Alcor secured packet); mirrored in tools/vs_reader.py and
covered by a new vector in tests/test_gsmtap.py.
- tools/capture_baseline.sh: add the alcor_reader_sample fixture; baseline
regenerated (APDU counts and RESULT: OK unchanged everywhere, 0 CONCAT).
- 75/75 unit tests. README/AGENTS updated.
v1.9.0 -- CLK frequency in RST events:
- RST events (sub_type 0x10) append the measured CLK frequency when known:
[direction, level, flags, clk_hz_be32], flags bit 0 = rate present; the
3-byte form is kept when the rate is not yet measured. Value is
samplerate / _samples_per_clock from the existing 16-edge measurement (PPS
changes F/D, not the CLK rate). VCC events unchanged. Backward
compatible (old consumers read the first two bytes only); consumers can
compute data_rate = clk_hz x D / F from the ATR/PPS Fi/Di.
- Unit tests for the payload codec; README/README_RUS/AGENTS updated.
v1.9.1 -- T=0 NULL is not an SW1:
- The 'scan for SW1' loops accepted any 0x6X byte, including the NULL
procedure byte 0x60 (card busy), so 'NULL + real SW1' was mis-framed as
the status word: the phone traces' SELECT MF came out as ...3f00 6061 —
real SW1 0x61 eaten as SW2, real SW2 0x29 left on the wire and discarded
as a suspicious CLA (desync cascade).
- plausible_sw() now accepts 0x61-0x6F and 0x90-0x9F only, per
ISO 7816-3:2006 10.3.3 Table 11 ('6X' or '9X', EXCEPT '60'; ISO 7816-4
enforces '60' as an invalid SW1) and UICC_SPECS.md 1.8. All seven
SW1-detection sites use plausible_sw(); NULLs met while scanning are
logged as rate-limited wait extensions (_note_pre_sw) instead of
suspicious bytes.
- tools/vs_reader.py::_plausible_sw() mirrors the fix; regression vector
for the 00a40004023f006061 artifact.
Validation: 74/74 unit tests; baseline regenerated and IDENTICAL (the SW
fix is count-neutral -- only the SW bytes change); all 9 example pcaps
regenerated -- vs_reader OK, zero APDUs ending in 60 61, SELECT MF now
61 29 / 61 25.
- Split the PPS exchange into the two standard GSMTAP-SIM subtypes:
0x02 (PPS_REQ, raw request bytes, span ss..es_req) and 0x03 (PPS_RSP,
raw response bytes, span es_req..end), replacing the single 0x02 packet
with concatenated request+response that deviated from libosmocore's
gsmtap.h. Wireshark now dissects both halves; simtrace2-pysniff needs
no change (its importer already maps 0x03 to 'pps' and _decode_pps()
parses the 0xFF-leading response frame). PPS annotation/log unchanged.
- gsmtap_stream.py: GSMTAP_SIM_PPS_REQ/RSP constants; header field labels
aligned with libosmocore's struct (signal_dbm/snr_db).
- Docs: GSMTAP event tables (EN/RU) updated, PPS deviation note removed;
corrected the res-byte flag attribution -- upstream simtrace2 never
sets res, so the GSMTAP_FLAG_BAD_FCS carriage is this decoder's own
extension (semantics mirror simtrace2's USB data flags).
- Tests: assert the PPS subtype values; baseline regenerated (only the
version line and test listing differed -- no decode-metric changes).
68/68 unit tests.
Residual CONCAT mis-frames (4gmodem/A55_sim2/S21p_sim2) traced to a silent
5-byte drop at an invalid-INS T=0 header (0x0f 00 a4 08 04). Drop only the
stray leading byte and re-frame the remaining four via _replay so the real
command (00 a4 08 04 02 2f 00 -> 61 1c) is recovered.
- _note_discard()/_consume_first(): every consumed-but-unemitted byte is
now logged; idle 0x00/0xFF discards are rate-limited, real data always
gets an ANN_WARN annotation.
- All read-until-SW loops log skipped non-SW bytes.
- All traces: CONCAT 0; BAD_FCS/GARBAGE 0; test_8/test_7 RESULT OK;
47/47 unit tests; baseline byte-identical.
History rewritten to a single root commit so no example trace is stored
in the repo (fixtures are local-only, gitignored). README keeps only
generic <trace>.sr placeholders; AGENTS.md retains the fixture names.
catarrh