read NTAG213 and NTAG216 instead of assuming NTAG215

[ccatlett1984]

fm175xx: read NTAG213 and NTAG216 instead of assuming NTAG215

Summary

__reader_a_ultralight_read_all_data is hard-coded to NTAG215's geometry. It loops to FM175XX_NTAG215_TOTAL_PAGES and allocates FM175XX_NTAG215_TOTAL_SIZE, so:

• NTAG216 is silently truncated to 540 of its 924 bytes. An NDEF record living in the upper 384 bytes is invisible, with no error to say so.
• NTAG213 fails outright. Reads past page 44 are NAKed, and the recovery branch cannot fire (see below), so the whole read returns FM175XX_CARD_READ_ERR.

It now reads until the tag stops answering and truncates to the largest recognised page count that was fully covered — 180 bytes for an NTAG213, 540 for an NTAG215, 924 for an NTAG216.

The recovery branch was dead code

if (page_no - 4) in Constants.FM175XX_ULTRALIGHT_VALID_END_PAGES:

FM175XX_ULTRALIGHT_VALID_END_PAGES is [135, 44], and page_no comes from range(0, 135, 4), so it takes values 0, 4, … 132 and page_no - 4 ranges over -4 … 128. It can never equal 135 (135 + 4 is not a multiple of 4) and never reaches 44 in a failing iteration, because the loop stops at 132 before a 44-page tag's first NAK at page 48 would be attempted. So any tag that NAKs mid-read falls through to FM175XX_CARD_READ_ERR, and any tag larger than an NTAG215 is quietly cut short.

This replaces the constant with FM175XX_ULTRALIGHT_KNOWN_PAGE_COUNTS and a check that can actually fire.

Why truncate rather than trust the read length

A READ returns four pages and rolls over within addressable memory, so the last successful read on any tag overruns the final page. Truncating to a known page count discards those roll-over bytes, which would otherwise be handed back as if they were tag memory.

Plain Ultralight is deliberately not in the recognised list, so it stays rejected as it was before. An Ultralight EV1 MF0UL21 has 41 pages and, because of roll-over, answers every read a 44-page tag would; accepting 44 would return 176 bytes whose tail is roll-over rather than memory. FM175XX_ULTRALIGHT_TOTAL_PAGES is left in place but is no longer referenced — happy to remove it if you would rather not keep an unused constant.

Testing

$ pytest
66 passed, 1 skipped

New test/test_fm175xx_ultralight_read.py drives the real function with the page-read helper replaced by a simulated tag that models both behaviours that matter — a READ returns four pages and rolls over within addressable memory, and an address past the last page is NAKed. No hardware is constructed or touched.

It asserts that an NTAG213, NTAG215 and NTAG216 each come back at their exact size and byte-identical to the simulated tag's memory — the roll-over check, since a naive implementation returns the right length with the wrong tail. It also asserts that 0-, 4-, 16-, 20-, 41- and 44-page tags are still rejected, 41 included specifically because it is a real size that roll-over makes look like 44.

Scope

Only the Ultralight/NTAG path changes. __reader_a_m1_read_all_data and the Mifare Classic path are untouched, as is every caller — read_mifare_ultralight still receives a byte list and simply gets a correctly sized one.

[ccatlett1984]

Closed due to a13bc9e