modules: add dirtydecrypt (CVE-2026-31635) + fragnesia (CVE-2026-46300)

Two new page-cache-write LPE modules, both ported from the public V12
security PoCs (github.com/v12-security/pocs):

- dirtydecrypt (CVE-2026-31635): rxgk missing-COW in-place decrypt.
  rxgk_decrypt_skb() decrypts spliced page-cache pages before the HMAC
  check, corrupting the page cache of a read-only file. Sibling of
  Copy Fail / Dirty Frag in the rxrpc subsystem.

- fragnesia (CVE-2026-46300): XFRM ESP-in-TCP skb_try_coalesce() loses
  the SHARED_FRAG marker, so the ESP-in-TCP receive path decrypts
  page-cache pages in place. A latent bug exposed by the Dirty Frag
  fix (f4c50a4034e6). Retires the old _stubs/fragnesia_TBD stub.

Both wrap the PoC exploit primitive in the skeletonkey_module
interface: detect/exploit/cleanup, an --active /tmp sentinel probe,
--no-shell support, and embedded auditd + sigma rules. The exploit
body runs in a forked child so the PoC's exit()/die() paths cannot
tear down the dispatcher. The fragnesia port drops the upstream PoC's
ANSI TUI (incompatible with a shared dispatcher); the exploit
mechanism is reproduced faithfully. Linux-only code is guarded with
#ifdef __linux__ so the modules still compile on non-Linux dev boxes.

VERIFICATION: ported, NOT yet validated end-to-end on a
vulnerable-kernel VM. The CVE fix commits are not pinned, so detect()
is precondition-only (PRECOND_FAIL / TEST_ERROR, never a blind
VULNERABLE) and --auto will not fire them unless --active confirms.
macOS stub-path compiles verified locally; the Linux exploit-path
build is covered by CI (build.yml, ubuntu) only. See each MODULE.md.

Wiring: core/registry.h, skeletonkey.c, Makefile, CVES.md, ROADMAP.md.

modules/dirtydecrypt_cve_2026_31635/MODULE.md

@@ -0,0 +1,77 @@
+# dirtydecrypt — CVE-2026-31635
+
+> 🟡 **PRIMITIVE / ported.** Faithful port of the public V12 PoC into
+> the `skeletonkey_module` interface. **Not yet validated end-to-end on
+> a vulnerable-kernel VM** — see _Verification status_ below.
+
+## Summary
+
+DirtyDecrypt (a.k.a. DirtyCBC) is a missing copy-on-write guard in
+`rxgk_decrypt_skb()` (`net/rxrpc/rxgk_common.h`). The function decrypts
+incoming rxgk socket buffers **in place** before the HMAC is verified.
+When the skb fragment pages are page-cache pages — spliced in via
+`MSG_SPLICE_PAGES` over loopback — the in-place AES decrypt corrupts the
+page cache of a read-only file.
+
+It is a sibling of Copy Fail (CVE-2026-31431) and Dirty Frag
+(CVE-2026-43284 / 43500): same bug class, different kernel subsystem
+(rxgk / AFS-style rxrpc encryption rather than algif_aead or xfrm-ESP).
+
+## Primitive
+
+Each `fire()`:
+
+1. Adds an `rxrpc` security key holding a crafted rxgk XDR token.
+2. Opens an `AF_RXRPC` client + a fake UDP server on loopback and
+   completes the rxgk handshake.
+3. Forges a DATA packet whose **wire header comes from userspace** and
+   whose **payload pages come from the target file's page cache**
+   (`splice` + `vmsplice`).
+4. The kernel decrypts the spliced page-cache pages in place — the HMAC
+   check then fails (expected), but the page cache is already mutated.
+
+`pagecache_write()` drives a **sliding-window** technique: byte[0] of
+each corrupted 16-byte AES block is uniformly random (≈1/256 chance of
+the wanted value), and round _i+1_ at offset _S+i+1_ overwrites the
+15-byte collateral of round _i_ without disturbing the byte round _i_
+fixed. Net cost ≈ 256 fires per byte.
+
+The exploit rewrites the first 120 bytes of a setuid-root binary
+(`/usr/bin/su` and friends) with a tiny ET_DYN ELF that calls
+`setuid(0)` + `execve("/bin/sh")`.
+
+## Operations
+
+| Op | Behaviour |
+|---|---|
+| `--scan` | Checks AF_RXRPC reachability + a readable setuid carrier. With `--active`, fires the primitive against a disposable `/tmp` file and reports VULNERABLE/OK empirically. |
+| `--exploit … --i-know` | Forks a child that corrupts the carrier's page cache and execs it for a root shell. `--no-shell` stops after the page-cache write. |
+| `--cleanup` | Evicts the carrier from the page cache (`POSIX_FADV_DONTNEED` + `drop_caches`). The on-disk binary is never written. |
+| `--detect-rules` | Emits embedded auditd + sigma rules. |
+
+## Preconditions
+
+- `AF_RXRPC` reachable (the `rxrpc` module loadable / built in).
+- A readable setuid-root binary to use as the payload carrier.
+- x86_64 (the embedded ELF payload is x86_64 shellcode).
+
+## Verification status
+
+This module is a **faithful port** of
+<https://github.com/v12-security/pocs/tree/main/dirtydecrypt>, compiled
+into the SKELETONKEY module interface. It has **not** been validated
+end-to-end against a known-vulnerable kernel inside the SKELETONKEY CI
+matrix.
+
+`detect()` deliberately does **not** return a kernel-version-based
+patched/vulnerable verdict: the CVE-2026-31635 fix commit is not yet
+pinned here, and fabricating a `kernel_patched_from` table would
+violate the project's no-fabrication rule (`CVES.md`). Instead:
+
+- preconditions missing → `PRECOND_FAIL`
+- preconditions present, no `--active` → `TEST_ERROR` ("cannot
+  determine passively") so `--auto` does not fire it blind
+- `--active` → empirical VULNERABLE / OK via the `/tmp` sentinel probe
+
+**Before promoting to 🟢:** pin the fix commit + branch-backport
+thresholds, add a `kernel_range`, and validate on a vulnerable VM.