SE050Sim: add the SE052F applet personality

[LinuxJedi]

Summary

Adds a fourth applet personality to SE050Sim for the SE052F (IoT applet
7.2.22), bench-characterized against real silicon. Selected with
SE050_SIM_APPLET=f / 52f / se052f / 7.2.22. The suffix rules are
checked before the leading-digit rule, so 3f selects the SE052F just as
3e selects the SE050E.

GetFreeMemory is U32 on this family

This is the substantive change, and it corrects the shape of an earlier
assumption rather than just adding data.

Se05x_API_GetFreeMemory picks its parser from SE05X_CHECK_52F_VERSION,
which tests (applet_version >> 8) & 0xFF for the range 0x10..=0x1F.
Because applet_version is major<<24 | minor<<16 | patch<<8, that byte is
the applet patch number, which NXP overloads to mark the SE052 family.
Applet 7.2.22 (patch 0x16) therefore takes the tlvGet_U32 branch, while
3.1.1 / 7.2.0 / SE050E (patch 0x00) take tlvGet_U16.

The two parsers reject each other's lengths outright, so a mismatched width
fails the host call rather than degrading quietly. free_memory_bytes now
carries u32 values internally and emits 2 or 4 bytes according to the new
free_memory_is_u32() predicate.

Memory type	SE052F
PERSISTENT	86336
TRANSIENT_RESET	1157
TRANSIENT_DESELECT	1152

PERSISTENT does not fit in a u16 at all, which is precisely why the U32
path exists. The other three personalities keep their 2-byte replies
unchanged.

Other modelled behaviour

All measured on silicon over Platform SCP03:

Behavior	SE051 7.2.0	SE052F 7.2.22
GetVersion blob	07 02 00 3f ff ff ff	07 02 16 26 f2 ff ff
GetFreeMemory width	2 bytes	4 bytes
GetRandom cap	1018	1003
ReadType, P-521 pair	0x31	0x31 (same)
Duplicate CreateECCurve	0x6985, curve intact	same
Ed25519 / X25519 keygen	works	0x6985
RSA-2048 CRT keygen + sign	works	works
RSA-1024 CRT keygen	works	0x6985
RSA public key import	works	0x6A80

appletConfig 0x26f2 clears EDDSA (0x0004), DH_MONT (0x0008), DES
(0x0100), MIFARE (0x0800) and RFU1 (0x1000) relative to the SE051's
0x3fff, while keeping both RSA bits. The RSA size and format restrictions
are not visible in appletConfig at all -- they only show up on the wire.

The GetRandom figure of 1003 was measured over SCP03, which is the only
channel a real SE052F serves: the part ships locked and refuses plain
GetRandom, so its plain-channel cap could not be measured. R-MAC/R-ENC
overhead is the likely reason it sits below the 1018 the other 7.2 parts
allow in plain mode.

handle_write_ec_key now takes the applet version so it can gate the 25519
curves. That is the only signature change.

Deliberately not modelled

• The factory restricted mode. Real SE052F parts refuse everything but
  GetVersion until an SCP03 session exists. Enabling that by default would
  make the personality unusable for anyone not driving SCP03. The mechanism
  already exists if it is ever wanted: the object store's scp_required
  flag plus SetPlatformSCPRequest.
• The P-256-only factory curve list. The simulator deliberately
  pre-provisions P-192/P-224/P-256/P-384/P-521 for every personality so
  hosts that never create curves keep working; the SE051 and SE050E factory
  lists differ from that default too, and neither is modelled.
• RSA key format. The SE052F refuses RAW-format generation with 0x6A80
  while accepting CRT, but the simulator does not distinguish the two
  formats for any personality.

Test plan

• cargo test -- --test-threads=1: 121 pass (93 lib + 14 + 14).
• New tests cover the GetFreeMemory reply width per personality (including
  the exact 00 01 51 40 byte pattern and that 86336 exceeds u16::MAX),
  the token parsing including the 3f / 3e precedence cases, the 25519
  refusals with the other personalities still serving those curves, and the
  RSA size and public-import restrictions.
• Integration suite (full T=1 -> APDU -> dispatch -> driver stack) run under
  each personality. Default and 3.1.1 are 14/14; SE050E shows its usual 4
  RSA failures, unchanged. Under SE050_SIM_APPLET=52f five tests fail, and
  each is the modelled silicon behaviour rather than a regression: all four
  RSA tests use 1024-bit keys, which a real SE052F refuses, and the fifth is
  Ed25519 keygen. test_get_free_memory passes under 52f, confirming the
  4-byte reply survives the whole stack.
• CI runs without SE050_SIM_APPLET, so the default personality is what is
  exercised there.

[Copilot]

Pull request overview

Adds a new SE050Sim “applet personality” for the SE052F (applet 7.2.22), including bench-validated behavioral differences (notably GetFreeMemory U32 width, 25519 curve disablement, and RSA restrictions) and corresponding unit tests.

Changes:

• Introduces AppletVersion::V7_2_22F and selection logic (SE050_SIM_APPLET token parsing, version bytes, feature predicates, GetRandom cap, GetFreeMemory width/values).
• Gates Ed25519/X25519 key handling based on applet version and models SE052F RSA restrictions (min key size + public-key import refusal).
• Expands tests across management/EC/RSA behavior to validate SE052F-specific wire behavior and selection precedence.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
SE050Sim/se050-sim/src/handlers/rsa.rs	Adds SE052F RSA restrictions + tests (min key size, public import refusal).
SE050Sim/se050-sim/src/handlers/management.rs	Updates tests for SE052F GetRandom cap, version bytes, and GetFreeMemory U32 replies.
SE050Sim/se050-sim/src/handlers/ec.rs	Adds applet-version gating to refuse 25519 curves on SE052F + tests; updates handler signature.
SE050Sim/se050-sim/src/handlers/curve.rs	Extends “duplicate CreateECCurve” behavior to include SE052F.
SE050Sim/se050-sim/src/dispatch.rs	Wires applet version through dispatch into EC handler.
SE050Sim/se050-sim/src/applet.rs	Adds SE052F personality metadata, selection parsing, feature predicates, and GetFreeMemory width/value logic + tests.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.