Skip to content

Run a virtual iPhone on an Apple Silicon Mac.

Swift

userland: Add DSC hv_vmm_present byte-5 mangle with sign-in blacklist and slot reattest (EXP only)

zqxwceaf90c9a903e6
diff --git a/research/0_binary_patch_comparison.md b/research/0_binary_patch_comparison.md
index 08d6d5b6..65d3ac6d 100644
--- a/research/0_binary_patch_comparison.md
+++ b/research/0_binary_patch_comparison.md
@@ -1,4 +1,30 @@
-# Patch Comparison: Regular / Development / Jailbreak
+# Patch Comparison: Regular / Development / Jailbreak / Experimental
+
+> **EXP is a JB superset.** Everything in the baseline tables below that is `Y`
+> for JB is also `Y` for EXP. The columns are kept at three variants to avoid
+> noise — the only place EXP and JB diverge is the **Experimental additions**
+> below, all of which are EXP-only (JB and the other variants are deliberately
+> unaffected). The EXP-only items, taken together:
+>
+> - **Kernel** — `KernelEXPPatcher` runs the `hv_vmm_present` sysctl OID rename
+> plus kernel-internal caller cstring/sandbox-profile-token mangle (formerly
+> wired into JB Group B as JB-26 — moved out).
+> - **DeviceTree at fw_patch time** — 8 identity-rewrite property patches
+> (Tier 1b + 1c) flipping userland-visible identity surfaces toward
+> D47AP / iPhone17,3.
+> - **DSC user-mode** — byte-5 cstring mangle of `kern.hv_vmm_present` with a
+> sign-in blacklist + per-page slot re-attestation (`cfw_patch_hv_vmm_dsc.py`),
+> companion to the kernel rename.
+> - **watchdogd (EXP-JB-3.5)** — surgical 2-instruction patch + slot re-attest;
+> forces the cached "am I a VM?" byte to `1` so watchdogd's clean-exit branch
+> runs.
+> - **Post-restore DT rewrite (EXP-JB-6)** — host-side rewrite of `devicetree.img4`
+> on the ramdisk's mounted rootfs for the three restore-fatal identity
+> properties (root `model`, `target-type`, `compatible[0]`) that broke
+> restore when applied at fw_patch time.
+> - **SystemVersion.plist `ProductBuildVersion` (EXP-JB-7, opt-in)** — gated on
+> `SPOOF_BUILD=<id>`. Rewrites the build identifier in the rootfs and
+> cryptex copies of `SystemVersion.plist`.
## Boot Chain Patches
@@ -103,6 +129,16 @@
| JB-24 | B | `patch_vm_fault_enter_prepare` | `_vm_fault_enter_prepare` | Force `cs_bypass` fast path in runtime fault validation | Y |
| JB-25 | B | `patch_vm_map_protect` | `_vm_map_protect` | Skip upstream write-downgrade gate in `vm_map_protect` | Y |
+### EXP-Only Kernel Methods (Reference List)
+
+Runs in `KernelEXPPatcher.findAll()` (chained after `KernelPatcher` +
+`KernelJBPatcher` for the `.exp` variant only — JB and other variants
+do NOT execute these).
+
+| # | Group | Method | Function | Purpose | EXP Enabled |
+| ------ | ----- | --------------------- | --------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :---------: |
+| EXP-01 | B | `patch_hv_vmm_rename` | sysctl OID name cstring `"hv_vmm_present"` → `"Xv_vmm_present"` (Part A) + every kernel-internal occurrence of `kern.hv_vmm_present` cstring/sandbox-profile token mangled at byte 5 (Part B) | Rename the `kern.hv_vmm_present` OID's name in place (`'h' → 'X'` at offset 0 of the 14-byte cstring). After this: `sysctlbyname("kern.hv_vmm_present")` returns ENOENT; `sysctlbyname("kern.Xv_vmm_present")` returns the original int value (1). Part B mangles every kernel-internal caller — AMFI, IOCryptoAcceleratorFamily, sandbox-profile token, apfs — so they keep hitting the renamed OID. Companion to the user-mode blacklist-flip mangle in `cfw_patch_hv_vmm_dsc.py`. | Y |
+
## CFW Installation Patches
### Binary Patches Applied Over SSH Ramdisk
@@ -115,6 +151,8 @@
| 4 | Plist injection | `launchd.plist` | bash/dropbear/trollvnc/vphoned daemons | Y | Y | Y |
| 5 | `b` (skip jetsam guard) | `launchd` | Prevent jetsam panic on boot | - | Y | Y |
| 6 | `LC_LOAD_DYLIB` injection | `launchd` | Load short alias `/b` (copy of `launchdhook.dylib`) at launch | - | - | Y |
+| 7 | cstring byte 5 mangle `'h' → 'X'` (`"kern.hv_vmm_present"` → `"kern.Xv_vmm_present"`) + per-page slot-hash re-attestation, BLACKLIST semantics — **EXP only** | DSC dylibs | Companion to EXP kernel rename (`KernelEXPPatcher.patchHvVmmRename`). The mangle is applied to every DSC dylib EXCEPT those in `DONT_PATCH_INSTALL_NAMES` (sign-in / device-likeness consumers, ~15 entries). Patched dylibs query `kern.Xv_vmm_present` and get the truthful 1 (graphics / accel passthrough). Blacklisted dylibs keep the original cstring, hit ENOENT on the renamed kernel, cache 0, lie about VM presence. On `codeSigningMonitor == 2` hardware the byte-mangle alone causes `CODESIGNING/Invalid Page` SIGKILL because TXM enforces per-page hashes; the re-attestation pass recomputes the SHA-256 slot in the chunk's `CS_CodeDirectory` for every modified 16 KiB page. See `scripts/patchers/cfw_dsc_codesign.py` and `cfw_patch_hv_vmm_dsc.py`. | - | - | - |
+| 8 | (removed — was: standalone-binary mangle in 6 rootfs Mach-Os via SSH) | n/a | Removed in the blacklist-flip redesign. With the EXP kernel rename in place, the 6 rootfs binaries (MobileActivationMigrator, CheckerBoard, StoreKitUISceneService, storekitd, appstored, CorePrescriptionService) get the desired "cache 0 / not in a VM" behavior for free: they keep their original cstring, hit ENOENT on the renamed kernel sysctl, defensive `cbnz w0, skip` leaves the cached byte at BSS-zero. No SSH-time standalone patch needed. | - | - | - |
### Installed Components
@@ -130,47 +168,385 @@
| 8 | BaseBin hooks | `systemhook.dylib` / `launchdhook.dylib` / `libellekit.dylib` -> `/cores/` plus `/b` alias for `launchdhook.dylib` | - | - | Y |
| 9 | `TweakLoader.dylib` | Lean user-tweak loader built from source and installed to `/var/jb/usr/lib/TweakLoader.dylib` | - | - | Y |
+### `kern.hv_vmm_present` user-mode patcher (EXP only)
+
+Companion to the EXP kernel patcher (`KernelEXPPatcher.patchHvVmmRename`).
+Mangles byte 5 of every `kern.hv_vmm_present` cstring inside DSC dylibs
+EXCEPT those in `DONT_PATCH_INSTALL_NAMES` (sign-in / device-likeness
+consumers, ~15 entries). Patched dylibs query the renamed OID and get the
+truthful 1 (graphics + accel passthrough); blacklisted dylibs keep the
+original cstring, hit ENOENT on the renamed kernel, and defensively cache 0
+("not running on a VM") for sign-in / device-attestation surfaces.
+Source-of-truth research: `research/hv_vmm_present_usermode_xrefs.md`.
+
+JB and other variants are NOT affected by this patcher.
+
+**Patch shape (every site)** — cstring mangle:
+
+```
+Before (cstring section bytes, 20 bytes total):
+ "kern.hv_vmm_present\0"
+ 6B 65 72 6E 2E 68 76 5F 76 6D 6D 5F 70 72 65 73 65 6E 74 00
+
+After (1 byte change at offset 0):
+ "Xern.hv_vmm_present\0"
+ 58 65 72 6E 2E 68 76 5F 76 6D 6D 5F 70 72 65 73 65 6E 74 00
+ ^^
+```
+
+The kernel's name-to-MIB translation fails with `ENOENT` when the
+caller asks for `"Xern.hv_vmm_present"`, so `sysctlbyname` returns
+-1. The canonical post-call check (`cbnz w0, skip` or
+`cmp w0,#0 ; b.ne skip`) then takes the skip-cache path; the cached
+"is_vmm" byte stays at its initial value (BSS-zero = 0).
+
+We don't modify executable code at all — only one byte of read-only
+string data. The kernel call still happens (with the wrong name), so
+any sysctl-tracing infrastructure can still see activity.
+
+Idempotent: a re-scan for the literal `"kern.hv_vmm_present\0"` finds
+no occurrences in already-mangled dylibs, so the patcher does no work
+on a re-run.
+
+**DSC-side patches** — driven by an explicit whitelist
+(`PATCH_INSTALL_NAMES` in `scripts/patchers/cfw_patch_hv_vmm_dsc.py`)
+applied to chunks under
+`SystemOS/System/Library/Caches/com.apple.dyld/`. Comment a line in
+the whitelist to skip that dylib on the next install — useful for
+bisecting which consumer is responsible for an observable change.
+
+| Dylib | Component role (paraphrased) |
+| --------------------------------------------------------- | ------------------------------------------------------------- |
+| `usr/lib/libMobileGestalt.dylib` | Backs `MGCopyAnswer("hv-vmm-present")` — highest fan-in |
+| `PrivateFrameworks/AAAFoundation.framework/AAAFoundation` | Apple ID anti-abuse plumbing |
+| `PrivateFrameworks/AuthKit.framework/AuthKit` | Sign-in-with-Apple-ID / iCloud auth |
+| `PrivateFrameworks/IDSFoundation.framework/IDSFoundation` | Apple Identity Service core (iMessage / FaceTime backbone) |
+| `PrivateFrameworks/DeviceIdentity.framework/DeviceIdentity` | Device-binding / device class identity |
+| `PrivateFrameworks/DeviceCheckInternal.framework/...` | DeviceCheck attestation |
+| `PrivateFrameworks/MobileActivation.framework/...` | Activation flow |
+| `PrivateFrameworks/ApplePushService.framework/...` | APNS client (claims device characteristics on connect) |
+| `PrivateFrameworks/AppStoreUtilities.framework/...` | Store / IAP support |
+| `PrivateFrameworks/CorePrescription.framework/...` | Health prescription store sync gate |
+| `PrivateFrameworks/CoreCDP.framework/CoreCDP` | CDP (cloud key-vault / iCloud Drive plumbing) |
+| `PrivateFrameworks/EmailFoundation.framework/...` | Mail account heuristics |
+| `PrivateFrameworks/PhotoFoundation.framework/...` | Photos asset visibility heuristics |
+| `PrivateFrameworks/FindMyBase.framework/FindMyBase` | Find My anti-spoof |
+| `PrivateFrameworks/AirPlaySupport.framework/...` | AirPlay receiver gate |
+| `PrivateFrameworks/TrialServer.framework/TrialServer` | A/B / trial-rollout exclude-VM gate |
+| `PrivateFrameworks/VisionKitCore.framework/VisionKitCore` | VisionKit |
+| `PrivateFrameworks/DVTInstrumentsUtilities.framework/...` | Xcode Instruments support |
+| `PrivateFrameworks/WatchdogServiceManagement.framework/...` | Watchdog manager |
+| `Frameworks/CoreVideo.framework/CoreVideo` | CoreVideo pipeline |
+
+**Standalone-binary patches (6 files, applied to the device rootfs
+over SSH)**
+
+| Path | Role |
+| ------------------------------------------------------------------------------------------------- | ------------------------------------------------------------- |
+| `/System/Library/DataClassMigrators/MobileActivationMigrator.migrator/MobileActivationMigrator` | Activation migration helper |
+| `/Applications/CheckerBoard.app/CheckerBoard` | Apple internal accessibility test app |
+| `/Applications/StoreKitUISceneService.app/StoreKitUISceneService` | StoreKit UI host |
+| `/System/Library/Frameworks/StoreKit.framework/Support/storekitd` | StoreKit / IAP daemon |
+| `/System/Library/PrivateFrameworks/AppStoreDaemon.framework/Support/appstored` | App Store daemon |
+| `/System/Library/PrivateFrameworks/CorePrescription.framework/XPCServices/CorePrescriptionService.xpc/CorePrescriptionService` | CorePrescription XPC service |
+
+**Explicitly NOT patched (compute / accel — patching here turns off
+VM fast-paths that exist so the lib doesn't try to touch real silicon
+ANE / AGX / hardware codecs):**
+
+```
+System/Library/Frameworks/CoreML.framework/CoreML
+System/Library/PrivateFrameworks/Espresso.framework/Espresso
+System/Library/PrivateFrameworks/AppleNeuralEngine.framework/AppleNeuralEngine
+System/Library/PrivateFrameworks/CoreRE.framework/CoreRE
+System/Library/PrivateFrameworks/RenderBox.framework/RenderBox
+System/Library/PrivateFrameworks/WebGPU.framework/WebGPU
+System/Library/PrivateFrameworks/caulk.framework/caulk
+System/Library/PrivateFrameworks/IOSurfaceAccelerator.framework/IOSurfaceAccelerator
+System/Library/ExtensionKit/Extensions/HostInferenceProviderService.appex/HostInferenceProviderService
+```
+
+**Wiring**
+
+* `scripts/patchers/cfw_patch_hv_vmm.py` — standalone cstring patcher
+ (used for the on-device files): finds the `"kern.hv_vmm_present\0"`
+ cstring in the Mach-O's __cstring section and rewrites its first
+ byte (`'k'` → `'X'`).
+* `scripts/patchers/cfw_dsc_chunks.py` — chunked-DSC byte-level helper
+ (`DSCChunks(chunks_dir)`): vmaddr↔chunk-fileoff mapping, cstring
+ scan over executable mappings, byte read/write at a vmaddr, and
+ Mach-O header walk-back to resolve a vmaddr to the dylib install
+ name (LC_ID_DYLIB).
+* `scripts/patchers/cfw_patch_hv_vmm_dsc.py` — DSC-native orchestrator.
+ No external `ipsw` dependency. For every `"kern.hv_vmm_present\0"`
+ occurrence in any executable mapping, walks back to the containing
+ dylib's Mach-O header, reads `LC_ID_DYLIB`, and — if the install
+ name is in the explicit `PATCH_INSTALL_NAMES` whitelist — rewrites
+ the first byte of the cstring through `DSCChunks.write_at_vma`.
+ Pure Python. Whitelist-based by design so an operator can comment
+ out individual entries to bisect.
+* `scripts/patchers/cfw.py patch-hv-vmm <binary>` —
+ standalone-Mach-O subcommand (used for the 6 on-device files).
+* `scripts/patchers/cfw.py patch-hv-vmm-dsc <chunks_dir>` —
+ DSC subcommand (used while the SystemOS Cryptex DMG is still
+ mounted on the host, before the device copy).
+* `scripts/patch_hv_vmm_userland.sh` — thin wrapper used by the
+ install scripts.
+* `scripts/cfw_install_exp.sh` — EXP install script. Pre-step before
+ invoking `cfw_install.sh`: decrypts the SysOS Cryptex into the cache
+ location `cfw_install.sh` already uses, mounts it, applies the
+ DSC patch, unmounts. The unmodified `cfw_install.sh` then sees the
+ cached (already-patched) DMG. Standalone watchdogd is patched later
+ via SSH at step `[EXP-JB-3.5]`.
+* `scripts/cfw_install_jb.sh` and `scripts/cfw_install_dev.sh` —
+ unchanged from pre-experimental baseline. Neither runs the DSC
+ patcher.
+
+### `kern.hv_vmm_present` kernel patcher — Part A + Part B (EXP only)
+
+The `KernelEXPPatchHvVmmRename` Swift patcher (in `KernelEXPPatcher.findAll()`,
+chained after `KernelPatcher` and `KernelJBPatcher` for the `.exp` variant only)
+renames the sysctl OID and rewrites every kernel-internal occurrence of the
+old name so kexts continue to find it under the new name. Two parts. JB and
+other variants do NOT run this patcher.
+
+**Part A — OID name rename.** Finds the OID's `oid_name` cstring as
+the NUL-delimited bytes `\0hv_vmm_present\0` (exactly one match
+required in the kernelcache; on iPhone17,3 / iOS 26.1 this lives at
+file offset `0x964e0` inside `com.apple.kernel`). Flips byte 0 of the
+cstring `'h'` (0x68) → `'X'` (0x58). After the patch, the kernel's
+`sysctl_register_oid` keeps the OID's MIB and value (1) intact but
+the name resolver returns `ENOENT` for `kern.hv_vmm_present` and
+returns 1 for `kern.Xv_vmm_present`.
+
+**Part B — kernel-internal caller mangle.** After Part A, any
+kernel-side `sysctlbyname("kern.hv_vmm_present", …)` call gets
+ENOENT and falls into the caller's "not in a VM" branch — which on
+the bring-up build caused AMFI to panic with `AMFI: No PMGR?
+(ConfigurationSettings.cpp:388)` during ramdisk boot. Part B mangles
+every kernel-internal occurrence of the `kern.hv_vmm_present` name
+so callers continue to find the renamed OID. The mangle flips byte
+5 of the inner cstring (`'h'` after `kern.`) → `'X'`, producing
+`kern.Xv_vmm_present`.
+
+Two byte-aligned forms are searched, both anchored at the
+`kern.hv_vmm_present` substring:
+
+| Form | Needle | Where it lives | Mangle delta within needle |
+|------|--------|----------------|----------------------------|
+| (i) NUL-delimited cstring | `\0kern.hv_vmm_present\0` | `__TEXT,__cstring` of any kext that calls sysctlbyname by full name | +6 (skip leading NUL + 5) |
+| (ii) Sandbox-profile name token | `kern.hv_vmm_present\x0f` | Inside a compiled sandbox-profile blob within `com.apple.security.sandbox`. The `\x0f` byte is the sandbox-profile end-of-name marker; the token has no leading NUL. | +5 |
+
+On iPhone17,3 / iOS 26.1 / 23B85 the universe is 5 occurrences
+(verified by raw substring scan over the kernelcache buffer):
+
+| File offset | Fileset entry | Form |
+|-------------|---------------|------|
+| `0x541d56` | `com.apple.driver.AppleMobileFileIntegrity` | (i) cstring |
+| `0x81bdc3` | `com.apple.iokit.IOCryptoAcceleratorFamily` | (i) cstring |
+| `0xa6618b` | `com.apple.security.sandbox` | (ii) sandbox-profile name token |
+| `0xbb0d55` | `com.apple.security.sandbox` | (i) cstring |
+| `0xbce1f9` | `com.apple.filesystems.apfs` | (i) cstring |
+
+Part B emits one patch record per match (5 total, plus Part A's 1)
+under patch IDs `kernelcache_exp.hv_vmm_internal_caller_mangle` and
+`kernelcache_exp.hv_vmm_oid_rename`. Idempotent: a re-run detects
+already-mangled bytes (`kern.Xv_vmm_present` instead of
+`kern.hv_vmm_present`) and reports the patch as already applied.
+
+**Note on the sandbox-profile occurrence.** This was missed by the
+original Part B because its needle required NUL on both sides. The
+sandbox-profile blob stores OID names as TLV-framed tokens where the
+trailing byte is `\x0f` (sandbox EOT) rather than a NUL. Without the
+second needle, sandboxed callers that interpret the profile's
+`kern.hv_vmm_present`-matching rule would still match against the
+OLD name, while the OID itself has been renamed — so the rule's
+ALLOW/DENY/audit action would never fire. With the second needle,
+the rule's name token is rewritten to `kern.Xv_vmm_present` and
+sandboxed callers that hit the renamed OID match the (rewritten)
+rule as intended. Covered occurrence verified on
+iPhone17,3 / iOS 26.1 / 23B85 at file offset `0xa6618b`.
+
+### `watchdogd` surgical hv_vmm_present cache patch (EXP only)
+
+**Why a dedicated patch.** After the EXP kernel-side OID rename
+(`KernelEXPPatchHvVmmRename`), `sysctlbyname("kern.hv_vmm_present", ...)`
+returns `ENOENT` on this image. `/usr/libexec/watchdogd` caches that
+answer at startup. On `ENOENT` the cached byte stays at its BSS-zero
+default (`0`) and the downstream `cbz w0, ...` at the IOWatchdog-lookup
+site (`+0x58e0`) takes a branch into a `_os_crash` wrapper that does
+`brk #1`. launchd's `_PanicOnCrash → PanicOnConsecutiveCrash = true`
+flag in `com.apple.watchdogd.plist` escalates the SIGTRAP to a kernel
+panic. The cstring-mangle approach used elsewhere doesn't apply here
+because we want this binary to behave as if the sysctl returned `1`,
+not as if it returned `ENOENT`.
+
+**Patch shape.** Two-instruction surgical edit at every site in
+watchdogd that has the canonical caching shape:
+
+```
+adrp x0, <page>
+add x0, x0, #<off> ; "kern.hv_vmm_present"
+...arg setup...
+bl _sysctlbyname
+cbnz w0, <skip> ; <-- patched: NOP
+ldur w8, [x29, #-4]
+cmp w8, #0
+cset wN, ne ; <-- patched: mov wN, #1
+adrp xM, <page>
+strb wN, [xM, #<imm>] ; cached "am I a VM?" byte
+```
+
+Net effect: the cached byte is forced to `1` regardless of the sysctl
+result, and watchdogd's pre-existing "detected virtual machine
+environment, exiting..." clean-exit branch runs instead of the trap
+path. Two functions in watchdogd match this shape on
+`iPhone17,3 / iOS 26.1`; both are patched.
+
+**Code signing.** The byte edit invalidates the SHA-256 slot hashes
+for the 4 KiB pages containing the modifications in watchdogd's own
+`CS_CodeDirectory`. The patcher recomputes those slot hashes in place
+via `cfw_macho_codesign.reattest_modified_offsets` (4 KiB page size
+read from the CD, correct tail-slot length, all present CDs). The
+resulting CD mutation also changes the cdHash, but the existing JB
+kernel patch `patch_amfi_cdhash_in_trustcache` accepts any cdHash, so
+AMFI's trust-cache check still passes at execve. The patcher does NOT
+re-sign with `ldid` — preserving the original Apple-issued code-signing
+identifier (`com.apple.watchdogd`) is required for launchd's boot-task
+identity validation; an earlier attempt to re-sign other rootfs
+binaries with `ldid_sign` tripped this check on `mobile_obliterator`.
+
+**Wiring.**
+
+* `scripts/patchers/cfw_macho_codesign.py` — standalone-Mach-O
+ page-hash re-attestation (parallel to `cfw_dsc_codesign.py` but
+ parses `LC_CODE_SIGNATURE` directly, uses page size from the CD
+ header, handles short tail slot, updates every present CD).
+* `scripts/patchers/cfw_patch_watchdogd.py` — capstone-anchored
+ pattern matcher + Keystone-assembled 2-insn patch + slot reattest.
+ Idempotent.
+* `scripts/patchers/cfw.py patch-watchdogd <binary>` — CLI subcommand.
+* `scripts/patch_hv_vmm_userland.sh watchdogd <binary>` — thin shim
+ used by the install script.
+* `scripts/cfw_install_exp.sh` — invokes the patcher at step `[EXP-JB-3.5]`
+ on the live `/mnt1/usr/libexec/watchdogd` (scp-down, patch, scp-up,
+ chmod 0755). JB and DEV install scripts do NOT run this step.
+
+### DeviceTree identity properties at fw_patch time (EXP only)
+
+`DeviceTreePatcher` carries two property-patch lists: `basePropertyPatches`
+(4 entries — `serial-number`, `home-button-type`, `artwork-device-subtype`,
+`island-notch-location`) applied for every variant, and
+`identityPropertyPatches` (8 entries) applied **only when `includeIdentityPatches`
+is true**, which `FirmwarePipeline` sets exactly when `variant == .exp`.
+
+The 8 EXP-only identity properties (no restore-fatal ones — those go through
+EXP-JB-6 post-restore):
+
+| # | Node path | Property | Old → New | Risk |
+|---|----------------------------------------------------|-----------------------|-------------------------------|---------|
+| 1 | `device-tree` | `target-sub-type` | `VPHONE600AP` → `D47AP` | HIGHER |
+| 2 | `device-tree` | `compatible[1]` | `iPhone99,11` → `iPhone17,3` (slot-preserving) | LOW |
+| 3 | `device-tree/product` | `fdr-product-type` | `iPhone99,11` → `iPhone17,3` | HIGHER |
+| 4 | `device-tree/product` | `sub-product-type` | `iPhone99,11` → `iPhone17,3` | LOW |
+| 5 | `device-tree/product` | `unique-model` | `VPHONE600AP` → `D47AP` | LOW |
+| 6 | `device-tree/arm-io` | `device_type` | `vresearch1-io` → `t8140-io` | MEDIUM |
+| 7 | `device-tree/arm-io` | `soc-generation` | `VResearch1` → `H17` | MEDIUM-LOW |
+| 8 | `device-tree/product/vphone600-gestalt-variants` | `name` (node rename) | `vphone600-gestalt-variants` → `d47-gestalt-variants` | LOW-MEDIUM |
+
+Root `model` and root `target-type` are deliberately NOT in this list —
+both have been empirically shown to break restore. Those edits run
+post-restore as EXP-JB-6.
+
+### Post-restore DT identity rewrite (EXP-JB-6, EXP only)
+
+After the restore daemon's BuildManifest identity check has passed,
+`cfw_install_exp.sh` step `[EXP-JB-6]` scp's `devicetree.img4` down from the
+mounted rootfs (`/mnt5/<boot-hash>/usr/standalone/firmware/`), runs
+`scripts/patchers/cfw_patch_post_restore_dt.py`, and scp's the rewritten
+img4 back. The Python patcher unwraps the IM4P via `pyimg4`, parses the
+DT flat-binary, rewrites three restore-fatal root properties, and repacks
+preserving the IMG4's original IM4M ticket. The iBSS/iBEC/LLB
+`image4_validate_property_callback` bypass (existing JB patch) accepts
+the modified payload at next boot.
+
+| # | Property | Old → New |
+|---|------------------|------------------------------------------------------------------------|
+| 1 | root `model` | `iPhone99,11` → `iPhone17,3` |
+| 2 | root `target-type` | `VPHONE600` → `D47` |
+| 3 | root `compatible` | reorder `["VPHONE600AP", "iPhone99,11", "AVP-ARM"]` → `["D47AP", "VPHONE600AP", "AVP-ARM"]` (keeps `VPHONE600AP` in second slot so IOKit's `AppleVMApple1IO` kext binding still resolves; userland reads only the first entry for `hw.model`) |
+
+Idempotent. Skipped if already-rewritten DT is detected.
+
+### SystemVersion.plist `ProductBuildVersion` rewrite (EXP-JB-7, EXP only, opt-in)
+
+Gated on the `SPOOF_BUILD` env var. When `cfw_install_exp.sh` is invoked
+with e.g. `SPOOF_BUILD=23F77`, step `[EXP-JB-7]` runs
+`scripts/patchers/cfw_patch_build_version.py` (plistlib-based,
+format-preserving) on both the rootfs and cryptex copies of
+`SystemVersion.plist` to rewrite the `ProductBuildVersion` key to the
+specified id. Without `SPOOF_BUILD`, the step is skipped and the build
+version stays at the original IPSW value.
+
+| File | Touched if `SPOOF_BUILD=<id>` |
+|-----------------------------------------------------------------------------------|:------------------------------:|
+| `/mnt1/System/Library/CoreServices/SystemVersion.plist` (rootfs) | Y |
+| `/mnt5/Cryptexes/OS/System/Library/CoreServices/SystemVersion.plist` (cryptex) | Y |
+
+`kern.osversion` is unaffected — that comes from a kernel global
+initialized from boot args, not from this plist. Userland MG cache
+picks up the new build identifier on first boot after the gestalt
+cache rebuild.
+
### CFW Installer Flow Matrix (Script-Level)
-| Flow Item | Regular (`cfw_install.sh`) | Dev (`cfw_install_dev.sh`) | JB (`cfw_install_jb.sh`) |
-| --------------------------------------------- | ------------------------------- | -------------------------- | --------------------------------------------- |
-| Base CFW phases (1/7 -> 7/7) | Runs directly | Runs directly | Runs via `CFW_SKIP_HALT=1 zsh cfw_install.sh` |
-| Dev overlay (`rpcserver_ios` replacement) | - | Y (`apply_dev_overlay`) | - |
-| SSH readiness wait before install | Y (`wait_for_device_ssh_ready`) | - | Y (inherited from base run) |
-| launchd jetsam patch (`patch-launchd-jetsam`) | - | Y (base-flow injection) | Y (JB-1) |
-| launchd dylib injection (`inject-dylib /b`) | - | - | Y (JB-1) |
-
-| Procursus bootstrap deployment | - | - | Y (JB-2) |
-| BaseBin hook deployment (`*.dylib` -> `/mnt1/cores`) | - | - | Y (JB-3) |
-| First-boot JB finalization (`vphone_jb_setup.sh`) | - | - | Y (post-boot; now fails before done marker if TrollStore Lite install does not complete) |
-| Additional input resources | `cfw_input` | `cfw_input` + `resources/cfw_dev/rpcserver_ios` | `cfw_input` + `cfw_jb_input` |
-| Extra tool requirement beyond base | - | - | `zstd` |
-| Halt behavior | Halts unless `CFW_SKIP_HALT=1` | Halts unless `CFW_SKIP_HALT=1` | Always halts after JB phases |
+| Flow Item | Regular (`cfw_install.sh`) | Dev (`cfw_install_dev.sh`) | JB (`cfw_install_jb.sh`) | EXP (`cfw_install_exp.sh`) |
+| --------------------------------------------- | ------------------------------- | -------------------------- | --------------------------------------------- | ----------------------------------------------------- |
+| Base CFW phases (1/7 -> 7/7) | Runs directly | Runs directly | Runs via `CFW_SKIP_HALT=1 zsh cfw_install.sh` | Runs via `CFW_SKIP_HALT=1 zsh cfw_install.sh` |
+| Dev overlay (`rpcserver_ios` replacement) | - | Y (`apply_dev_overlay`) | - | - |
+| SSH readiness wait before install | Y (`wait_for_device_ssh_ready`) | - | Y (inherited from base run) | Y (inherited from base run) |
+| launchd jetsam patch (`patch-launchd-jetsam`) | - | Y (base-flow injection) | Y (JB-1) | Y (JB-1) |
+| launchd dylib injection (`inject-dylib /b`) | - | - | Y (JB-1) | Y (JB-1) |
+| Procursus bootstrap deployment | - | - | Y (JB-2) | Y (JB-2) |
+| BaseBin hook deployment (`*.dylib` -> `/mnt1/cores`) | - | - | Y (JB-3) | Y (JB-3) |
+| First-boot JB finalization (`vphone_jb_setup.sh`) | - | - | Y (post-boot) | Y (post-boot) |
+| DSC pre-patch (`kern.hv_vmm_present` byte-5 mangle + slot reattest) | - | - | - | Y (pre-step, before base CFW) |
+| `watchdogd` surgical 2-insn patch + slot reattest | - | - | - | Y (EXP-JB-3.5) |
+| Post-restore DT identity rewrite (`devicetree.img4`)| - | - | - | Y (EXP-JB-6) |
+| `SystemVersion.plist` `ProductBuildVersion` rewrite | - | - | - | Y (EXP-JB-7, opt-in via `SPOOF_BUILD`) |
+| Additional input resources | `cfw_input` | `cfw_input` + `resources/cfw_dev/rpcserver_ios` | `cfw_input` + `cfw_jb_input` | `cfw_input` + `cfw_jb_input` |
+| Extra tool requirement beyond base | - | - | `zstd` | `zstd` |
+| Halt behavior | Halts unless `CFW_SKIP_HALT=1` | Halts unless `CFW_SKIP_HALT=1` | Always halts after JB phases | Always halts after EXP phases |
## Summary
-| Component | Regular | Dev | JB |
-| ------------------------ | ------: | --: | --: |
-| AVPBooter | 1 | 1 | 1 |
-| iBSS | 2 | 2 | 3 |
-| iBEC | 3 | 3 | 3 |
-| LLB | 6 | 6 | 6 |
-| TXM | 1 | 12 | 12 |
-| Kernel (base) | 28 | 29 | 28 |
-| Kernel (JB methods) | - | - | 59 |
-| Boot chain total | 41 | 53 | 112 |
-| CFW binary patches | 4 | 5 | 6 |
-| CFW installed components | 6 | 7 | 9 |
-| CFW total | 10 | 12 | 15 |
-| Grand total | 51 | 65 | 127 |
+| Component | Regular | Dev | JB | EXP |
+| ---------------------------------- | ------: | --: | --: | --: |
+| AVPBooter | 1 | 1 | 1 | 1 |
+| iBSS | 2 | 2 | 3 | 3 |
+| iBEC | 3 | 3 | 3 | 3 |
+| LLB | 6 | 6 | 6 | 6 |
+| TXM | 1 | 12 | 12 | 12 |
+| Kernel (base) | 28 | 29 | 28 | 28 |
+| Kernel (JB methods) | - | - | 59 | 59 |
+| Kernel (EXP methods, `hv_vmm`) | - | - | - | 6 |
+| DeviceTree base properties | 4 | 4 | 4 | 4 |
+| DeviceTree EXP identity properties | - | - | - | 8 |
+| Boot chain total | 45 | 57 | 116 | 130 |
+| CFW binary patches (base) | 4 | 5 | 6 | 6 |
+| CFW EXP-only steps | - | - | - | 4 (DSC pre-patch, watchdogd EXP-JB-3.5, post-restore DT EXP-JB-6, build-version EXP-JB-7 opt-in) |
+| CFW installed components | 6 | 7 | 9 | 9 |
+| CFW total | 10 | 12 | 15 | 19 |
+| Grand total | 55 | 69 | 131 | 149 |
## Ramdisk Variant Matrix
-| Variant | Pre-step | `Ramdisk/txm.img4` | `Ramdisk/krnl.ramdisk.img4` | `Ramdisk/krnl.img4` | Effective kernel used by `ramdisk_send.sh` |
-| ------------- | ------------------- | -------------------------------- | -------------------------------------------------------------------------------- | ----------------------------------------- | --------------------------------------------------- |
-| `RAMDISK` | `make fw_patch` | release TXM + base TXM patch (1) | base kernel (28), legacy `*.ramdisk` preferred else derive from pristine CloudOS | restore kernel from `fw_patch` (28) | `krnl.ramdisk.img4` preferred, fallback `krnl.img4` |
-| `DEV+RAMDISK` | `make fw_patch_dev` | release TXM + base TXM patch (1) | base kernel (28), same derivation rule | restore kernel from `fw_patch_dev` (29) | `krnl.ramdisk.img4` preferred, fallback `krnl.img4` |
-| `JB+RAMDISK` | `make fw_patch_jb` | release TXM + base TXM patch (1) | base kernel (28), same derivation rule | restore kernel from `fw_patch_jb` (28+59) | `krnl.ramdisk.img4` preferred, fallback `krnl.img4` |
+| Variant | Pre-step | `Ramdisk/txm.img4` | `Ramdisk/krnl.ramdisk.img4` | `Ramdisk/krnl.img4` | Effective kernel used by `ramdisk_send.sh` |
+| -------------- | -------------------- | -------------------------------- | -------------------------------------------------------------------------------- | ---------------------------------------------- | --------------------------------------------------- |
+| `RAMDISK` | `make fw_patch` | release TXM + base TXM patch (1) | base kernel (28), legacy `*.ramdisk` preferred else derive from pristine CloudOS | restore kernel from `fw_patch` (28) | `krnl.ramdisk.img4` preferred, fallback `krnl.img4` |
+| `DEV+RAMDISK` | `make fw_patch_dev` | release TXM + base TXM patch (1) | base kernel (28), same derivation rule | restore kernel from `fw_patch_dev` (29) | `krnl.ramdisk.img4` preferred, fallback `krnl.img4` |
+| `JB+RAMDISK` | `make fw_patch_jb` | release TXM + base TXM patch (1) | base kernel (28), same derivation rule | restore kernel from `fw_patch_jb` (28+59) | `krnl.ramdisk.img4` preferred, fallback `krnl.img4` |
+| `EXP+RAMDISK` | `make fw_patch_exp` | release TXM + base TXM patch (1) | base kernel (28), same derivation rule | restore kernel from `fw_patch_exp` (28+59+6) | `krnl.ramdisk.img4` preferred, fallback `krnl.img4` |
## Cross-Version Dynamic Snapshot
diff --git a/research/firmware_manifest_and_origins.md b/research/firmware_manifest_and_origins.md
index 7c1074e6..aac06b6d 100644
--- a/research/firmware_manifest_and_origins.md
+++ b/research/firmware_manifest_and_origins.md
@@ -318,3 +318,257 @@ The firmware is a PCC shell wrapping an iPhone core. The vresearch101 boot chain
handles DFU/TSS signing. The vphone600 device tree + SEP + kernel provide the
runtime environment. The iPhone userland is patched post-install for activation
bypass, jailbreak tools, and persistent SSH/VNC.
+
+---
+
+## 9. DeviceTree identity rewrite
+
+The vphone600ap DT carries 13 properties whose values contain the literals
+`vphone600` / `VPHONE600` / `iPhone99,11` / `vmapple` / `vresearch`. Most are
+load-bearing (kernel kext-matching, restore-time identity, IORegistry path
+structure) and cannot be rewritten without breaking boot or restore.
+`DeviceTreePatcher` rewrites the 8 that are either userland-facing identity
+fields or SoC-generation descriptors with non-binding semantics.
+
+| # | Path | Property | Original | Patched | Slot | Risk |
+|----|------------------------------------------------------|----------------------|-----------------------------------------------------------|----------------------------------------------------------|------|---------------|
+| 2 | `device-tree` | `target-sub-type` | `"VPHONE600AP"` | `"D47AP"` | 12 | HIGHER |
+| 3 | `device-tree` | `compatible[1]` | secondary `"iPhone99,11"` (multi-string) | `"iPhone17,3"` (multi-string, primary `"VPHONE600AP"` preserved) | 48 | LOW |
+| 10 | `device-tree/product` | `fdr-product-type` | `"iPhone99,11"` | `"iPhone17,3"` | 12 | HIGHER |
+| 11 | `device-tree/product` | `sub-product-type` | `"iPhone99,11"` | `"iPhone17,3"` | 12 | LOW |
+| 12 | `device-tree/product` | `unique-model` | `"VPHONE600AP"` | `"D47AP"` | 12 | LOW |
+| 6 | `device-tree/arm-io` | `device_type` | `"vresearch1-io"` | `"t8140-io"` (matches d47ap) | 14 | MEDIUM |
+| 7 | `device-tree/arm-io` | `soc-generation` | `"VResearch1"` | `"H17"` (matches d47ap) | 11 | MEDIUM-LOW |
+| 13 | `device-tree/product/vphone600-gestalt-variants` | `name` (node name) | `"vphone600-gestalt-variants"` | `"d47-gestalt-variants"` | 27 | LOW-MEDIUM |
+
+**Properties intentionally left at vphone600 values** (with the empirical
+reason for each):
+
+- `device-tree.target-type` (`"VPHONE600"`) — Tier 2 attempt broke restore;
+ iBoot / restored_external read this and reject the device when the value
+ doesn't match the BuildManifest's signed identity.
+- `device-tree.model` (`"iPhone99,11"`) — Tier 1 attempt broke restore;
+ same reason as `target-type`.
+- `device-tree.compatible[0]` (`"VPHONE600AP"`) — IOKit's platform-expert
+ matching at boot binds against the FIRST entry of `compatible`. Rewriting
+ it to `D47AP` forces `_PE_init_platform_expert` to look for a D47AP-claiming
+ kext, which doesn't exist in our kernelcache. Panic guaranteed.
+- `device-tree/arm-io.compatible` / `device_type` / `soc-generation`
+ (vmapple1 / vresearch1-io / VResearch1) — bind the vmapple1 IO controller
+ kext, GIC, and PCIe drivers. Rename → no kext claims the SoC IO → boot
+ panic.
+- `device-tree/arm-io/gic.compatible`, `device-tree/arm-io/pcie.compatible` —
+ same family as arm-io entries.
+- `device-tree/product/vphone600-gestalt-variants` (node name) — referenced
+ by libMobileGestalt-equivalent code that looks up the subtree by literal
+ node name. Rename breaks the lookup.
+
+**Compatible multi-string surgical rewrite** (#3). The original blob is:
+
+```
+[0..10] "VPHONE600AP" (11 bytes)
+[11] NUL
+[12..22] "iPhone99,11" (11 bytes)
+[23] NUL
+[24..46] "AppleVirtualPlatformARM" (23 bytes)
+[47] NUL
+```
+
+Patched (slot length preserved at 48):
+
+```
+[0..10] "VPHONE600AP" ← unchanged: platform-expert bind site
+[11] NUL
+[12..21] "iPhone17,3"
+[22] NUL
+[23..45] "AppleVirtualPlatformARM"
+[46] NUL
+[47] NUL (pad)
+```
+
+`AppleVirtualPlatformARM` shifts 1 byte earlier but every consumer walks
+the blob by NUL-terminators, none depend on a fixed byte offset within
+the property. IOKit's platform-expert match against `VPHONE600AP` (first
+entry) is unaffected; the second entry, which userland walks of the
+compatible[] list see when enumerating identifiers, now reports
+`iPhone17,3`.
+
+**Risk classification.**
+
+- **LOW** (#3, #11, #12) — read only by userland identity APIs
+ (libMobileGestalt, IORegistry queries, App Store device class).
+ No iBoot / restored_external / TSS dependency.
+- **HIGHER** (#2, #10) — `target-sub-type` is in the same restore-signed
+ family as `target-type` (already proven to break restore);
+ `fdr-product-type` is read by Factory-Data-Restore code paths that may
+ check it against the BuildManifest. If restore fails after a build that
+ enables these, remove them first and try the remaining LOW-risk three
+ in isolation.
+
+Wiring: all 5 patches are declared in
+`sources/FirmwarePatcher/DeviceTree/DeviceTreePatcher.swift` under
+`propertyPatches`. The serializer rebuilds the entire tree on every run,
+so each new value is written at the same slot length as the original.
+The `compatible` patch uses a new `PropertyValue.bytes(Data)` case that
+takes a pre-built byte blob (necessary because the multi-string layout
+contains embedded NULs).
+
+## 10. Post-restore DT identity rewrite (EXP-JB-6)
+
+The properties under "known-bad from prior rounds" (root `model`, root
+`target-type`) cannot be edited at fw_patch time because
+`restored_external` / iBoot's restore mode cross-checks them against the
+BuildManifest's signed `SupportedProductTypes`. But that cross-check
+fires **only during installation**. After restore completes, the boot
+path is
+
+```
+AVPBooter (patched)
+ → iBSS / iBEC (image4_validate_property_callback bypass patches)
+ → LLB (image4 bypass + rootfs bypass patches)
+ → iBoot (research) → DT loaded from disk
+ → kernel → sysctls populated from DT
+```
+
+The image4 bypass patches we already ship are what allow every
+fw_patch-time IM4P modification to boot at all — they make the IM4M's
+stored digest irrelevant on subsequent boots. So an additional IM4P
+modification applied **after** restore completes is in-policy with the
+project's existing trust-chain bypass.
+
+The **EXP-JB-6** install step exploits this (in `cfw_install_exp.sh`
+only — JB and DEV variants do NOT run it). While the install pipeline
+still has `/mnt5` (the preboot volume) mounted in ramdisk mode, it:
+
+1. Pulls `/mnt5/<boot-hash>/usr/standalone/firmware/devicetree.img4` to
+ the host.
+2. Runs `scripts/patchers/cfw_patch_post_restore_dt.py`, which:
+ - Unwraps img4 → IM4P → decompresses LZFSE → DT blob (via pyimg4).
+ - Applies the three restore-unsafe edits (table below).
+ - Re-compresses LZFSE → repacks IM4P → repacks IMG4 with the original
+ IM4M.
+3. Pushes the modified img4 back to the same path.
+4. The device reboots out of ramdisk; iBoot loads the modified DT;
+ kernel populates `machine_info` from the new property values.
+
+| Property | Before (post-Tier-1b state on disk) | After (EXP-JB-6 rewrites to) | Slot |
+|---------------------------|---------------------------------------------------------------------------|-------------------------------------------------------------------------|------|
+| `device-tree.model` | `"iPhone99,11"` | `"iPhone17,3"` | 12 |
+| `device-tree.target-type` | `"VPHONE600"` | `"D47"` | 10 |
+| `device-tree.compatible` | `["VPHONE600AP", "iPhone17,3", "AppleVirtualPlatformARM"]` (post-Tier-1b) | `["D47AP", "VPHONE600AP", "AppleVirtualPlatformARM"]` (reordered) | 48 |
+
+The `compatible` rewrite is a **reorder**, not a replacement: VPHONE600AP
+stays in the list (now as the second entry), so IOKit's platform-expert
+match against the AppleVMApple1IO kext still binds. `compatible[0]`
+becomes D47AP, which is what userland's `hw.model` resolves to.
+
+Userland-visible effects after the next boot:
+
+- `sysctl hw.machine` → `"iPhone17,3"` (was `"iPhone99,11"`)
+- `sysctl hw.product` → `"iPhone17,3"` (was `"iPhone99,11"`)
+- `sysctl hw.model` → `"D47AP"` (was `"VPHONE600AP"`)
+- Settings → General → About → Model Identifier reflects the new
+ ProductType.
+- Every `MGCopyAnswer` key backed by `IOPlatformExpertDevice` properties
+ picks up the new values.
+
+Idempotence: the patcher reads the existing DT, applies edits, and
+writes back only if there's a difference. A second run on an
+already-patched devicetree.img4 detects target-state-already-met and
+exits without rewriting.
+
+Wiring:
+
+* `scripts/patchers/cfw_patch_post_restore_dt.py` — host-side patcher.
+ Uses pyimg4 for img4↔IM4P+LZFSE round-trip; mirrors the DT format
+ parser/serializer from `DeviceTreePatcher.swift`.
+* `scripts/cfw_install_exp.sh [EXP-JB-6]` — install step that runs after
+ JB-5 (LaunchDaemon deploy) and before the CLEANUP/umount block, in the
+ EXP install script only. Discovers the boot-manifest-hash via the same
+ `get_boot_manifest_hash` helper used by earlier install steps. Tolerates
+ a missing devicetree.img4 with warn-and-skip rather than aborting the
+ install. JB and DEV install scripts do NOT carry this step.
+
+Why this is restore-safe (and the fw_patch-time version was not):
+
+- **fw_patch-time path** modifies the IM4P bytes that are sent during
+ restore. iBoot's restore mode reads each IM4P and (via the image4
+ validation callback the iBSS/iBEC/LLB patches already neuter) accepts
+ its digest mismatch — but some other path inside `restored_external`
+ enforces the top-level `SupportedProductTypes` check independently and
+ rejects DT.model values that don't appear in that list.
+- **EXP-JB-6 post-restore path** never re-runs the BuildManifest check.
+ Subsequent boots validate IM4P contents against the IM4M only — which
+ the image4 bypass patches already wave through. So the same byte
+ change that's fatal at restore is harmless after restore.
+
+## 11. Build-version rewrite (EXP-JB-7, opt-in)
+
+**Opt-in step**. EXP-JB-7 runs only when the `SPOOF_BUILD` environment
+variable is set — e.g. `make setup_machine EXP=1 SPOOF_BUILD=23F77`, or
+`make cfw_install_exp SPOOF_BUILD=23F77` if you're re-running just the
+CFW phase. When omitted/empty the step is skipped entirely and the
+build identifier stays at whatever the IPSW shipped. Only the EXP
+install script runs this step; JB and DEV variants are not affected.
+
+The iPhone IPSW we install from ships with build identifier `23B85` (iOS
+26.1). iOS displays this string in Settings → General → About → "Build"
+and exposes it through `MGCopyAnswer("BuildVersion")`, `CoreFoundation`'s
+`_CFCopyServerVersionDictionary`, App Store telemetry, and every other
+framework path that reads `/System/Library/CoreServices/SystemVersion.plist`.
+
+The build identifier lives in exactly two on-device plist files. Both
+are plain XML/binary plists (no Apple-side per-file signature), and
+both live on volumes that are writable at install time:
+
+| Path | Volume | Role |
+|-------------------------------------------------------------------------------------|--------------|----------------------------------------------------------------------|
+| `/System/Library/CoreServices/SystemVersion.plist` | rootfs | The canonical source. Read by every userland identity API. |
+| `/private/preboot/Cryptexes/OS/System/Library/CoreServices/SystemVersion.plist` | preboot | Cryptex-side copy. Read by Cryptex-aware OS-version queries. |
+
+The **EXP-JB-7** install step rewrites the `ProductBuildVersion` key in both
+plists to a target value (currently `23F77`):
+
+```
+ProductBuildVersion: "23B85" → "23F77"
+```
+
+`ProductVersion` (`26.1`), `ProductName` (`iPhone OS`), `BuildID`,
+`SystemImageID`, and `ProductCopyright` are left untouched.
+
+Wiring:
+
+* `scripts/patchers/cfw_patch_build_version.py` — host-side plistlib-based
+ rewriter. Auto-detects XML vs binary plist format and preserves it on
+ write. Idempotent — a re-run on an already-patched plist exits without
+ rewriting.
+* `scripts/cfw_install_exp.sh [EXP-JB-7]` — install step that runs after
+ EXP-JB-6 (post-restore DT) and before CLEANUP, in the EXP install script
+ only. For each of the two plist paths it scp_from's the file to host,
+ runs the patcher with target `23F77`, scp_to's the file back. Tolerates
+ missing-on-device with warn+continue. JB and DEV install scripts do NOT
+ carry this step.
+
+What this **does not** flip:
+
+- `sysctl kern.osversion` — populated at boot from a kernel global
+ initialized from boot args, not from `SystemVersion.plist`. The kernel
+ image we ship was built against `23B78` (the PCC vphone600 /
+ vresearch101 build identifier; visible in 621 kext-Info.plist
+ embedded `DTPlatformBuild` / `DTSDKBuild` blobs inside the kernelcache
+ but never read by `kern.osversion` at runtime). To flip
+ `kern.osversion` we'd need to either rebuild the kernelcache with a
+ different `OS_BUILD_VERSION` or patch the kernel boot args path — out
+ of scope for this round.
+- `SystemVersionCompat.plist` — carries `23B34a` / `19.1`, a legacy
+ iOS-19 marker for `MacCatalyst`-style version queries. Not user-
+ visible, deliberately untouched.
+- DSC dylib internals — the apparent `23B85` byte sequence in DSC chunk
+ `.21` is a Swift mangled-name UUID-uniquing substring
+ (`_TtC11MediaCoreUIP33_98519F523B8515A67EEFBCB0824D82807Counter`), not
+ a build identifier. Patching it would corrupt a Swift symbol name.
+
+Order of operations: this step runs **after** EXP-JB-6 (the DT identity
+rewrite) so that any post-restore identity work targeting `/mnt5` has
+completed before we modify the Cryptex's SystemVersion.plist (which
+lives on the same volume).
diff --git a/research/hv_vmm_present_usermode_xrefs.md b/research/hv_vmm_present_usermode_xrefs.md
new file mode 100644
index 00000000..8ff68121
--- /dev/null
+++ b/research/hv_vmm_present_usermode_xrefs.md
@@ -0,0 +1,487 @@
+# `kern.hv_vmm_present` user-mode call sites — iPhone17,3 26.1 (23B85)
+
+Source material: `ipsws/iPhone17,3_26.1_23B85_Restore` (raw IPSW; encrypted DMGs)
+and `ipsws/iPhone17,3_26.1_23B85_Restore_extracted` (rebuilt rootfs with split DSC).
+
+This document enumerates every place in user space that reads
+`kern.hv_vmm_present` via the `sysctl` family on this firmware, what each
+caller appears to do with the result, and what is reasonable to expect from
+patching that specific reader instead of the kernel.
+
+## Current shipping design — blacklist-flip + kernel rename
+
+The current patch pipeline takes a **kernel-rename + user-mode-blacklist**
+approach. Summary:
+
+* The Swift kernel patch `patch_hv_vmm_rename` (JB-26, Group B) renames
+ the kernel's `hv_vmm_present` sysctl OID name in place to
+ `Xv_vmm_present` (single byte change at offset 0 of the 14-byte
+ cstring `\0hv_vmm_present\0`). After this kernel patch:
+ - `sysctlbyname("kern.hv_vmm_present", ...)` returns ENOENT.
+ - `sysctlbyname("kern.Xv_vmm_present", ...)` returns 1, the OID's
+ real int value.
+* The user-mode mangle pattern moved from byte 0 (`'k' → 'X'`,
+ producing `Xern.hv_vmm_present` — unroutable because `Xern` isn't a
+ top-level namespace) to byte 5 (`'h' → 'X'`, producing
+ `kern.Xv_vmm_present` — keeps `kern.` intact, routes to the renamed
+ OID).
+* `scripts/patchers/cfw_patch_hv_vmm_dsc.py` now applies the mangle as
+ a BLACKLIST. The list `DONT_PATCH_INSTALL_NAMES` names dylibs whose
+ cstring is LEFT untouched — they query the original (now
+ ENOENT-returning) name and cache 0 ("not in a VM"). Every other DSC
+ dylib with the cstring gets the byte-5 mangle, queries the new name,
+ and caches 1 ("in a VM", same as stock) — keeping graphics +
+ compute/accel fast paths intact.
+* The standalone-binary patch step over SSH (originally `cfw_install_jb.sh`
+ JB-3.5 and `cfw_install_dev.sh` 6.5/7) was removed. With the kernel
+ rename in place, those 6 rootfs binaries fall into the "unpatched → ENOENT
+ → cache 0" bucket automatically.
+
+The earlier whitelist-only design (no kernel patch, mangle a chosen
+subset to ENOENT) is preserved for reference in the section
+"Pre-blacklist whitelist enumeration" below, since the per-caller
+classification work is still the source-of-truth for what each
+binary's cstring usage actually looks like.
+
+## What the sysctl is
+
+`kern.hv_vmm_present` is defined in XNU at
+`bsd/kern/hvg_sysctl.c:176`:
+
+> `static SYSCTL_PROC(_kern, OID_AUTO, hv_vmm_present,`
+> ` CTLTYPE_INT | CTLFLAG_ANYBODY | CTLFLAG_KERN | CTLFLAG_LOCKED, …)`
+
+Backing handler `sysctl_vmm_present` (line 136 in the same file) returns
+an `int`, computed on arm64 from `IODTGetDefault("vmm-present", …)` —
+i.e. a device-tree key that Apple's Virtualization.framework /
+PCC bring-up plants in the guest. On real hardware the key is absent
+and the read returns `0`; on PCC/research VMs it returns `1`. That is
+exactly the path you are flipping today by patching the kernel.
+
+Because this is a single named OID with `CTLFLAG_ANYBODY`, every user-mode
+reader funnels through the canonical `sysctlbyname()` shape:
+
+```
+adrp x0, <page> ; x0 = "kern.hv_vmm_present"
+add x0, x0, #<off>
+sub x1, x29, #4 ; x1 = &result (int)
+mov x2, sp ; x2 = &len (= 4)
+mov x3, #0 ; newp
+mov x4, #0 ; newlen
+bl _sysctlbyname
+```
+
+This shape is the matcher I used to flag “real readers” below.
+
+## Method
+
+1. Located the literal `"kern.hv_vmm_present\0"` inside every Mach-O of the
+ rebuilt rootfs (the split DSC dylibs and every executable). 56 binaries
+ matched on bytes alone.
+2. For each match, parsed the LC\_SEGMENT\_64 layout and located the
+ actual section. Anything that wasn't in `__TEXT,__cstring` (or another
+ section flagged `S_CSTRING_LITERALS`, flag `0x2`) was investigated
+ separately.
+3. For each cstring hit, looked for ADRP+ADD or ADRP+LDR pairs in any
+ executable section that compute the string's vm address, then walked
+ forward up to ~24 instructions tracking simple register propagation
+ (`mov xN, xM`).
+4. Classified each xref:
+ * `CANONICAL` — string ends up in `x0`, `x3=x4=0`, BL within the window.
+ Matches `sysctlbyname(name, oldp, oldlenp, NULL, 0)`.
+ * `PROBABLE` — string in `x0` and a BL in window, but `x3`/`x4` zeroing
+ not seen.
+ * `REFERENCED` — xref exists but it is loaded into a non-x0 register or
+ used as an address anchor for adjacent data, not handed to a syscall.
+ * `STORED` — string written into a struct (i.e. a registry, not a
+ direct caller).
+ * `DEAD` — string lives in `__cstring` but no executable instruction
+ points at its page.
+
+The full per-binary results (snippets, addresses, classification) are
+saved at `outputs/hv_vmm_present_xref.json`.
+
+## Inventory by class
+
+### A. CANONICAL `sysctlbyname("kern.hv_vmm_present", …)` callers — 35
+
+Each line is `<binary> string@<vmaddr> xref@<vmaddr>`. The xref
+address is the **`add` instruction** that finalizes the pointer value
+in `x0`; immediately followed by the canonical 5-arg `sysctlbyname`
+preamble and a `bl`.
+
+```
+usr/lib/libMobileGestalt.dylib string@0x1b0201c60 xref@0x1b0198d00
+System/Library/Frameworks/CoreVideo.framework/CoreVideo string@0x19d17a4bb xref@0x19d122eb8
+System/Library/Frameworks/CoreML.framework/CoreML string@0x1950ed0d0 xref@0x194c3c754
+System/Library/Frameworks/StoreKit.framework/Support/storekitd string@0x10031a890 xref@0x100176f6c
+System/Library/PrivateFrameworks/CoreRE.framework/CoreRE string@0x1e3171b20 xref@0x1e2790f28
+System/Library/PrivateFrameworks/Espresso.framework/Espresso string@0x1949204e5 xref@0x193bef83c
+System/Library/PrivateFrameworks/AppleNeuralEngine.framework/AppleNeuralEngine string@0x1ad297375 xref@0x1ad248248
+System/Library/PrivateFrameworks/AppStoreDaemon.framework/Support/appstored string@0x10046dae6 xref@0x1003e25b8
+System/Library/PrivateFrameworks/AppStoreUtilities.framework/AppStoreUtilities string@0x23ff916d7 xref@0x23ff8bfe4
+System/Library/PrivateFrameworks/ApplePushService.framework/ApplePushService string@0x1b235fd50 xref@0x1b2358d10
+System/Library/PrivateFrameworks/AuthKit.framework/AuthKit string@0x1933c11de xref@0x19322f7e8
+System/Library/PrivateFrameworks/AAAFoundation.framework/AAAFoundation string@0x1c86581a1 xref@0x1c8650ee4
+System/Library/PrivateFrameworks/AirPlaySupport.framework/AirPlaySupport string@0x2223618a0 xref@0x222306a28
+System/Library/PrivateFrameworks/CorePrescription.framework/CorePrescription string@0x2471fe6ed xref@0x2471ebc18
+System/Library/PrivateFrameworks/CorePrescription.framework/XPCServices/CorePrescriptionService.xpc/CorePrescriptionService string@0x100092175 xref@0x100005064
+System/Library/PrivateFrameworks/CoreCDP.framework/CoreCDP string@0x1dedfa62d xref@0x1dedbb724
+System/Library/PrivateFrameworks/DVTInstrumentsUtilities.framework/DVTInstrumentsUtilities string@0x247f42ad0 xref@0x247f1154c
+System/Library/PrivateFrameworks/DeviceCheckInternal.framework/DeviceCheckInternal string@0x2487a018c xref@0x24878d10c
+System/Library/PrivateFrameworks/DeviceIdentity.framework/DeviceIdentity string@0x2260a4b25 xref@0x2260862a4
+System/Library/PrivateFrameworks/EmailFoundation.framework/EmailFoundation string@0x1c61cd1f7 xref@0x1c6165e88
+System/Library/PrivateFrameworks/FindMyBase.framework/FindMyBase string@0x1df635420 xref@0x1df5fb8d4
+System/Library/PrivateFrameworks/IDSFoundation.framework/IDSFoundation string@0x1a7ea0fe4 xref@0x1a7c65dc4
+System/Library/PrivateFrameworks/IOSurfaceAccelerator.framework/IOSurfaceAccelerator string@0x2547cd71c xref@0x2547caa48
+System/Library/PrivateFrameworks/MobileActivation.framework/MobileActivation string@0x2598f06c6 xref@0x2598e2618
+System/Library/PrivateFrameworks/PhotoFoundation.framework/PhotoFoundation string@0x1d8ba907c xref@0x1d8b9e7fc
+System/Library/PrivateFrameworks/RenderBox.framework/RenderBox string@0x195e54555 xref@0x195d344a0
+System/Library/PrivateFrameworks/TrialServer.framework/TrialServer string@0x26f32471e xref@0x26f224120
+System/Library/PrivateFrameworks/VisionKitCore.framework/VisionKitCore string@0x1b442d9cf xref@0x1b435f0a4
+System/Library/PrivateFrameworks/WatchdogServiceManagement.framework/WatchdogServiceManagement string@0x272725b23 xref@0x272725114
+System/Library/PrivateFrameworks/WebGPU.framework/WebGPU string@0x225722b6c xref@0x22552c2bc
+System/Library/PrivateFrameworks/caulk.framework/caulk string@0x27512204f xref@0x275112808
+System/Library/DataClassMigrators/MobileActivationMigrator.migrator/MobileActivationMigrator string@0x72aa xref@0x3024
+System/Library/ExtensionKit/Extensions/HostInferenceProviderService.appex/HostInferenceProviderService string@0x1000112b0 xref@0x10000b890
+Applications/CheckerBoard.app/CheckerBoard string@0x10006fb56 xref@0x1000192e8
+Applications/StoreKitUISceneService.app/StoreKitUISceneService string@0x1000a2ca0 xref@0x100083f04
+```
+
+All 35 readers share the same five-argument shape. Most also share a
+post-call idiom: store the boolean result into a static byte (cache),
+and a sibling `getter()` reads the cached byte. libMobileGestalt is the
+clearest exemplar:
+
+```
+0x1b0198cf0 stur wzr, [x29, #-4]
+0x1b0198cf4 mov w8, #4
+0x1b0198cf8 str x8, [sp]
+0x1b0198cfc adrp x0, #0x1b0201000
+0x1b0198d00 add x0, x0, #0xc60 ; "kern.hv_vmm_present"
+0x1b0198d04 sub x1, x29, #4
+0x1b0198d08 mov x2, sp
+0x1b0198d0c mov x3, #0
+0x1b0198d10 mov x4, #0
+0x1b0198d14 bl #<sysctlbyname>
+0x1b0198d18 cbnz w0, ...skip...
+0x1b0198d1c ldur w8, [x29, #-4]
+0x1b0198d24 cset w8, ne
+0x1b0198d28 adrp x9, #0x1ed446000
+0x1b0198d2c strb w8, [x9, #0xcf8] ; cached_is_vmm := (val != 0)
+```
+
+CoreML, CoreRE, Espresso, AppleNeuralEngine, MobileActivation,
+caulk, RenderBox, etc. all do the same: one call, cached, sibling
+accessor. That makes the patch surface for each one very small —
+overwriting `cset w8, ne` with `mov w8, #0` (or NOPing the call and
+zeroing the cached byte's accessor) is all that's needed per binary.
+
+### B. False positives ruled out by xref analysis
+
+15 of the 56 string-bearing binaries are **not** actual readers:
+
+#### B.1 Embedded entitlement plist (the binary just declares it can read the OID)
+
+These four embed a `com.apple.security.exception.sysctl.read-only` (or
+`sysctl-read`) entitlement that names `kern.hv_vmm_present`. The
+entitlement only authorises the sandbox; the actual `sysctlbyname` call
+happens in libraries they link (almost certainly libMobileGestalt or
+CoreFoundation). The binaries themselves contain no instruction
+xref to the string.
+
+```
+Applications/Family.app/Family
+Applications/PeopleMessageService.app/PeopleMessageService
+Applications/PeopleViewService.app/PeopleViewService
+System/Library/PrivateFrameworks/DataDetectorsUI.framework/PlugIns/com.apple.DataDetectorsUI.ActionsExtension.appex/com.apple.DataDetectorsUI.ActionsExtension
+```
+
+#### B.2 Compiled sandbox profile (the string lives inside an embedded SBPL blob)
+
+The string is in `__TEXT,__const`, surrounded by high-entropy bytes that
+match a compiled sandbox profile. No code reaches the string. As with
+B.1, these only declare the OID is permitted — not that they read it.
+
+```
+usr/sbin/absd
+usr/sbin/fairplayd.H2
+System/Library/CoreServices/ClarityBoard.app/ClarityBoard
+System/Library/PrivateFrameworks/CoreALD.framework/CoreALD
+System/Library/PrivateFrameworks/IDS.framework/identityservicesd.app/identityservicesd
+System/Library/PrivateFrameworks/SonicKit.framework/SonicKit
+System/Library/PrivateFrameworks/VideosUI.framework/VideosUI
+```
+
+#### B.3 String present in `__cstring` but unreferenced (linker DCE leftover)
+
+The string sits in the cstring section but no ADRP+ADD/LDR in any
+executable section computes its vmaddr, and no 8-byte little-endian
+encoding of the address (raw, image-relative, or 36-bit chained-fixup
+shape) appears in any other section either.
+
+```
+usr/sbin/bluetoothd (verified: 0 ADRPs into the page; 0 byte-pointer matches)
+System/Library/Frameworks/MediaToolbox.framework/MediaToolbox
+System/Library/Frameworks/ManagedAppDistribution.framework/Support/managedappdistributiond
+System/Library/PrivateFrameworks/ApplePushService.framework/apsd
+System/Library/PrivateFrameworks/CMCapture.framework/CMCapture
+System/Library/PrivateFrameworks/MagnifierSupport.framework/MagnifierSupport
+System/Library/PrivateFrameworks/NeuralNetworks.framework/NeuralNetworks
+System/Library/PrivateFrameworks/Recon3D.framework/Recon3D
+System/Library/PrivateFrameworks/VFX.framework/VFX
+```
+
+#### B.4 Adjacent‑string anchor (REFERENCED but not a sysctl call)
+
+```
+System/Library/Frameworks/SoundAnalysis.framework/SoundAnalysis
+```
+
+The ADRP+ADD here computes `x8 = "kern.hv_vmm_present"`, then immediately
+`x20 = x8 - 0x20`, i.e. it is using our string as a fixed offset to reach
+a *different* sysctl name 32 bytes earlier in the cstring section. The
+sysctlbyname call passes `x20`, not our string. No need to patch.
+
+## What each CANONICAL caller is reasonable to expect to do
+
+I am being literal here: I report what the framework path implies and
+what the disassembly shows. Beyond what's listed I do not have direct
+evidence of behaviour and would need runtime tracing to be sure.
+
+| Component | Likely role of the check | Patching effect (only this caller) |
+|---|---|---|
+| `libMobileGestalt.dylib` | Implements `MGCopyAnswer("hv-vmm-present")` and similar — the answer most other Apple frameworks query through it. Single shared cache byte. | Many higher-level callers ask MobileGestalt instead of doing their own sysctl. Forcing MG's cache to 0 makes most “am I a VM?” checks throughout the system see "no" without further code changes. **Highest-leverage single patch.** |
+| `CoreML.framework` / `Espresso.framework` / `AppleNeuralEngine.framework` / `CoreRE.framework` / `RenderBox.framework` / `WebGPU.framework` / `caulk.framework` | Compute / accel paths. Each independently caches a "running on VM" flag; their respective dispatchers gate ANE/Metal/HW codecs. | Disables their VM-aware fast-path bypasses. Practical effect: lets these libraries try the same code path they'd use on real silicon. On a research VM with PCC PV=3 they may now hit code that demands real ANE/SEP/etc. and fail — patch only if you specifically need them to take the silicon path. |
+| `IOSurfaceAccelerator.framework` | Hardware surface allocation paths. | Same caveat — pushing it onto the silicon path may produce IOKit failures. |
+| `RenderBox.framework`, `WebGPU.framework`, `Espresso.framework` | Already covered above (graphics/ML). | — |
+| `CoreVideo.framework` | Video pipeline / capture surfaces. Caches a global. | Similar story to the accel libs. |
+| `StoreKit.framework` (`storekitd`), `AppStoreDaemon` (`appstored`), `AppStoreUtilities`, `Applications/StoreKitUISceneService` | All of the App Store / IAP daemons and their host UI. They tag receipts with VM status. | Best candidate to patch *together*, because making the Store stop labelling sessions as "vm=1" while everything else stays the same is exactly the kind of selective lying the user asked for. Receipts and SKUI flow start looking like a real device. |
+| `ApplePushService` (`ApplePushService`) | APNS client. APNS connection establishment includes a "device characteristics" payload. | Patching makes the APNS handshake claim the device is bare metal. Useful if push tokens are silently downgraded for VMs. |
+| `AAAFoundation`, `AuthKit`, `IDSFoundation`, `DeviceIdentity`, `DeviceCheckInternal` | Apple ID / iCloud / iMessage attestation and DeviceCheck (anti-fraud) plumbing. Every one of them caches `is_vmm` to feed into device-binding/anti-abuse decisions. | Patching the AuthKit/IDS/DeviceCheck callers individually is the conservative way to make “Sign in with Apple ID”, iMessage activation, and DeviceCheck attestations stop self-flagging as VM, while leaving compute paths (CoreML, ANE) alone. **High-leverage for "make iMessage/iCloud trust this device" without disabling kernel-level VM behaviour.** |
+| `MobileActivation.framework`, `MobileActivationMigrator` | Activation flow + migration tool. | Activation today uses kern.hv_vmm_present to decide whether to take the fastlane / dev path. Patching would normalise activation. |
+| `CorePrescription.framework` (+ XPC service) | Health-data prescription store. | The check is almost certainly "don't sync prescriptions on a VM"; patching here gates Health/Rx data sync. |
+| `EmailFoundation.framework` | Mail. | Likely a mail-account heuristic; patching is benign. |
+| `PhotoFoundation.framework` | Photos. | Likely a "hide some assets on VMs" check. |
+| `FindMyBase.framework` | Find My. | Anti-spoof gate. |
+| `TrialServer.framework` | Internal A/B / trial-rollout client. | Trials' "exclude VMs" gate. |
+| `WatchdogServiceManagement` | The watchdog manager. | Note `string@0x272725b23` lives in the same page as its own `__text`; this is a tiny binary with one and only one consumer. Patching is local and self-contained. |
+| `VisionKitCore.framework`, `CoreCDP.framework` | Vision/CDP. Single sysctl, cached. | — |
+| `AirPlaySupport.framework` | AirPlay. Cached flag. | Likely a "no AirPlay receiver on VM" gate. |
+| `DVTInstrumentsUtilities.framework` | Xcode Instruments support library. | Dev-only; cosmetic. |
+| `Applications/CheckerBoard.app` | Apple's internal CoreAccessibility test app — the build-time sample most likely link-imports the same telemetry helper as everyone else. | Cosmetic. |
+| `HostInferenceProviderService.appex` | On-device generative-AI inference host. | Same family as the ML accel libs above; gate to compute paths. |
+
+## Why the kernel-side patch breaks display while a user-side patch does not
+
+The kernel-side patch (returning `0` from `sysctl_vmm_present` /
+`IODTGetDefault("vmm-present")`) flips the answer for **every** consumer
+in the boot pipeline, including kernel-internal users of `vmm-present`
+(IOKit, IOSurface, display drivers, AGX). The display path consumes the
+flag for its own paravirt routing — that's the symptom you observed.
+
+A user-mode patch only changes what landed in the 35 dylib/exe call sites
+above. None of them are on the display bring-up path; they all run
+post-launchd, after the framebuffer/IOSurface plumbing is already up.
+So the display works as VM, while the higher-level "this is a VM"
+hints are suppressed wherever you choose.
+
+## Recommended patch surface
+
+Given the vphone-cli design intent ("VM-aware where it has to be,
+device-like where it can be"), the smallest set that buys the most is:
+
+1. **`libMobileGestalt.dylib`** — single biggest fan-in. Many other
+ processes ask MG instead of calling `sysctlbyname` directly, so this
+ one patch propagates.
+2. **`CorePrescription`, `WatchdogServiceManagement`, `EmailFoundation`,
+ `PhotoFoundation`, `FindMyBase`, `AirPlaySupport`, `TrialServer`,
+ `CoreCDP`, `VisionKitCore`** — the consumer apps/services. Patching
+ them gives the device-like surface for sync, mail, photos, find-my,
+ etc.
+3. **`AAAFoundation`, `AuthKit`, `IDSFoundation`, `DeviceIdentity`,
+ `DeviceCheckInternal`, `MobileActivation`, `MobileActivationMigrator`,
+ `ApplePushService`** — identity / push / activation. Patching here
+ normalises the device's posture against Apple's anti-abuse signals.
+4. **`StoreKit/storekitd`, `AppStoreDaemon/appstored`, `AppStoreUtilities`,
+ `StoreKitUISceneService`** — App Store + IAP. Patch as a group.
+5. **DO NOT patch (most likely):** `CoreML`, `CoreRE`, `Espresso`,
+ `AppleNeuralEngine`, `RenderBox`, `WebGPU`, `caulk`, `CoreVideo`,
+ `IOSurfaceAccelerator`, `HostInferenceProviderService`. These are the
+ compute/accel libraries — pushing them onto the silicon path on a VM
+ is the most likely way to cause new failures (no real ANE / no real
+ AGX / no real H.265 hardware encoder available to take the call).
+6. Ignore the false positives in section B entirely — they don't read
+ the sysctl.
+
+The exact byte-level patch shape that works for every CANONICAL site is
+identical to what you already have in the kernel: turn the 1-bit boolean
+that gets cached after the call into a constant 0. The ARM64 idiom
+`cset wN, ne` (after `cmp wM, #0`) → `mov wN, #0` (3 byte-pattern
+substitutions: `0x9f1f00b1` zero immediate move, etc.) is a one-instruction
+patch per consumer.
+
+If you'd rather skip the call entirely, replace the `bl <sysctlbyname>`
+with `mov w0, #0` (success) and zero the result on the stack — also
+one instruction, but now `w0=0` is read as the int result so the
+post-call boolean naturally becomes 0.
+
+## Files
+
+* `outputs/hv_vmm_present_xref.json` — full per-binary xref dump
+ (string addresses, xref addresses, classification, surrounding
+ disassembly).
+* `research/hv_vmm_present_xref.json` — same dump committed in-tree.
+
+## Patcher implementation (Dev + JB only)
+
+Implemented as part of the existing `cfw_install_dev.sh` /
+`cfw_install_jb.sh` flows; the regular `cfw_install.sh` is left
+untouched (per design — the regular variant stays "as honest as
+possible", and the device-likeness layer is opt-in via the dev or
+jb variants).
+
+**Patch shape, every site** — 1-byte cstring mangle:
+
+```
+cstring "kern.hv_vmm_present\0" → "Xern.hv_vmm_present\0"
+ ^ byte 0 = 'k' (0x6B) ^ byte 0 = 'X' (0x58)
+```
+
+The kernel returns `ENOENT` for the mangled name; the canonical
+post-call error check (`cbnz w0, skip` / `cmp w0,#0 ; b.ne skip`)
+takes the skip-cache path, so the cached "is_vmm" byte stays at its
+initial 0 (stack pre-zeroed by `stur wzr` or BSS-zero for a global
+`oldp`). Idempotent: a re-run can't find the unmangled cstring and
+no-ops.
+
+We don't touch executable code — only one byte of string data in
+`__TEXT,__cstring`. The kernel call still happens (so sysctl
+instrumentation still sees activity), it just fails the
+name-to-MIB translation.
+
+**What gets patched**
+
+| Tier | Scope | Where applied |
+| ---------------- | -------------------- | --------------------------------------------------------- |
+| DSC dylibs (20) | Identity / store / consumer-services (see list above) | DSC chunks inside the SystemOS Cryptex DMG, while it is mounted on the host |
+| Standalone (6) | MobileActivationMigrator, CheckerBoard, StoreKitUISceneService, storekitd, appstored, CorePrescriptionService.xpc | Pulled from the device rootfs over SSH, patched, ldid-signed, pushed back |
+| Compute / accel (10) | CoreML, Espresso, ANE, CoreRE, RenderBox, WebGPU, caulk, IOSurfaceAccelerator, HostInferenceProviderService | **Not patched** — leaving these alone preserves their VM-aware fast-path bypasses |
+
+**Source files**
+
+* `scripts/patchers/cfw_patch_hv_vmm.py` — standalone cstring patcher
+ (`patch_hv_vmm(filepath)`): finds occurrences of `"kern.hv_vmm_present\0"`
+ in `__cstring` (and the objc method/class name pools, for safety) and
+ rewrites byte 0 from `'k'` to `'X'`.
+* `scripts/patchers/cfw_dsc_chunks.py` — chunked-DSC byte-level helper:
+ vmaddr ↔ chunk-fileoff mapping, executable-mapping cstring scan,
+ byte read/write at a vmaddr, and Mach-O header walk-back to resolve
+ a vmaddr to the containing dylib's install name.
+* `scripts/patchers/cfw_patch_hv_vmm_dsc.py` — DSC-native orchestrator.
+ No external `ipsw` dependency. For every `"kern.hv_vmm_present\0"`
+ occurrence found in any executable mapping, walks back to the
+ containing dylib's Mach-O header, reads `LC_ID_DYLIB`, and — if the
+ install name is in the explicit `PATCH_INSTALL_NAMES` whitelist —
+ rewrites the first byte of the cstring through
+ `DSCChunks.write_at_vma`. After all whitelisted byte-mangles, calls
+ `cfw_dsc_codesign.reattest_modified_pages` to recompute the SHA-256
+ slot hash in the containing DSC chunk's `CS_CodeDirectory` for every
+ affected 16 KiB page (see below — this is required on hardware with
+ `codeSigningMonitor == 2`). The whitelist is one entry per line in
+ the source file so an operator can comment out individual dylibs
+ during bisection. (An earlier draft used `bl _sysctlbyname` →
+ `movn w0,#0` instruction patching and an `ipsw dyld extract` path;
+ both were scrapped — the former because it modifies executable
+ bytes which has the same per-page-hash risk as the cstring approach
+ but a larger surface area, and the latter because `ipsw` returns a
+ Mach-O whose load-command file offsets reference the original DSC
+ chunk's file layout, making it unparseable as a standalone Mach-O.)
+* `scripts/patchers/cfw_dsc_codesign.py` — page-hash re-attestation.
+ Without this pass, every byte-mangle invalidates the affected 16 KiB
+ page's SHA-256 slot hash in the chunk-level `CS_CodeDirectory`. On
+ iPhone17,3 / iOS 26.1 (`codeSigningMonitor == 2`), TXM enforces
+ per-page integrity at demand-page-in (XNU defers exec-page
+ validation to TXM for `csm_associated` mappings —
+ `research/reference/xnu/osfmk/vm/vm_fault.c:2763-2780`). Concrete
+ repro before re-attestation: `lsd-2026-05-11-035240.ips` faults
+ inside MobileActivation's __TEXT at offset `0x13DC0`, in the same
+ 16 KiB page as the mangled cstring (offset `0x106C6`). The
+ re-attestation rewrites only the 32-byte slot whose page contains
+ a modified byte; the CDHash side-effect (CD blob hash changes
+ when slots change) is accepted by TXM at DSC mount time
+ (empirically verified — device boots with re-attested DSC).
+* `scripts/patchers/cfw.py` — adds `patch-hv-vmm <binary>` and
+ `patch-hv-vmm-dsc <chunks_dir>` subcommands.
+* `scripts/patch_hv_vmm_userland.sh` — thin wrapper used by the
+ install scripts.
+* `scripts/cfw_install_dev.sh` — DSC patch is applied while the
+ SystemOS Cryptex DMG is still mounted on the host (inside step
+ `[1/7]`); standalone binaries are patched over SSH in new step
+ `[6.5/7]`.
+* `scripts/cfw_install_jb.sh` — pre-step before invoking the
+ regular `cfw_install.sh`: decrypts the SysOS Cryptex into the
+ same cache file `cfw_install.sh` would, mounts it, applies the
+ DSC patch, unmounts. The unmodified `cfw_install.sh` then sees
+ the cached (already-patched) DMG. Standalone binaries are patched
+ in new step `[EXP-JB-3.5]` after the base install completes (EXP variant only).
+
+## Dylibs that must keep seeing VM=1 (graphics + accel passthrough)
+
+Empirically validated: lying to these dylibs about VM presence breaks
+the boot graphics path or the compute/accel fast paths. They MUST
+read 1 (or the original truthful value) at runtime to function. Under
+the new blacklist-flip design they are therefore NOT in
+`DONT_PATCH_INSTALL_NAMES` — which means the byte-5 mangle IS applied
+to their cstrings, they query the renamed kernel sysctl
+`kern.Xv_vmm_present`, and the kernel returns 1 to them just like a
+stock device would.
+
+| Dylib | Why a lie breaks it |
+| -------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `/usr/lib/libMobileGestalt.dylib` | Highest fan-in answer source. Forcing it to think "not a VM" sends the whole display-init chain into paths that demand real silicon display hardware. |
+| `PrivateFrameworks/PhotoFoundation.framework/PhotoFoundation` | Photos / image pipeline ties into the boot display path (asset decode + thumbnail rendering touches the same Metal surfaces the boot UI uses). |
+| `PrivateFrameworks/AirPlaySupport.framework/AirPlaySupport` | AirPlay screen-mirroring plumbing registers as a display source at boot; lying about VM mode flips the registration path. |
+| `PrivateFrameworks/VisionKitCore.framework/VisionKitCore` | Camera / vision-intelligence stack hooks into the boot-time display chain on this device. |
+| `Frameworks/CoreVideo.framework/CoreVideo` | Directly in the display pipeline — lying routes the boot display path through code that assumes physical capture surfaces. |
+| `Frameworks/CoreML.framework/CoreML` | Compute fast-path. The VM-aware bypass is what prevents CoreML from trying to drive real ANE. |
+| `PrivateFrameworks/Espresso.framework/Espresso` | Same — neural-net inference dispatch. |
+| `PrivateFrameworks/AppleNeuralEngine.framework/AppleNeuralEngine` | ANE driver bring-up. |
+| `PrivateFrameworks/CoreRE.framework/CoreRE` | Reality Engine dispatch. |
+| `PrivateFrameworks/RenderBox.framework/RenderBox` | Graphics dispatch. |
+| `PrivateFrameworks/WebGPU.framework/WebGPU` | GPU dispatch. |
+| `PrivateFrameworks/caulk.framework/caulk` | HW codec dispatch. |
+| `PrivateFrameworks/IOSurfaceAccelerator.framework/IOSurfaceAccelerator` | IOSurface accelerator dispatch. |
+| `ExtensionKit/Extensions/HostInferenceProviderService.appex/HostInferenceProviderService` | On-device ML inference host. |
+
+Under the new design these libs are not enumerated in code — they
+implicitly fall into "not blacklisted → patched → sees VM=1" by
+virtue of NOT being in `DONT_PATCH_INSTALL_NAMES`. To explicitly
+exclude a different lib from the patch (force it to lie about VM
+presence), add its install name to the blacklist. The build / install
+flow is idempotent in re-attestation (`cfw_patch_hv_vmm_dsc.py` will
+detect already-mangled cstrings on a re-run and re-sync slot hashes
+for their pages), so toggling an entry in or out of the blacklist
+doesn't require a fresh IPSW unpack — but doing so after a previous
+install requires `rm -f vm/.cfw_temp/CryptexSystemOS.dmg` to force a
+fresh DSC mount.
+
+## What is verified vs. inferred
+
+Verified by direct disassembly + Mach-O byte inspection:
+* The 35 CANONICAL sites use `sysctlbyname("kern.hv_vmm_present", &int, &len, NULL, 0)`.
+* The 4 entitlement-plist-only matches are pure plist substring hits.
+* The 7 sandbox-profile-only matches are inside a compiled SBPL blob
+ (string lives in `__TEXT,__const`, no executable xref, identical
+ byte signature across the seven binaries).
+* The 9 DEAD-string matches really are unreferenced (no ADRP/LDR
+ computes the address; no 8-byte little-endian or 36-bit chained-fixup
+ shape contains it).
+* `SoundAnalysis` uses our string only as an offset anchor.
+
+Inferred from framework / binary path and the canonical caching shape
+(not from runtime tracing): the mapping in the table above of "what
+each caller uses the result for". The disassembly proves *that* they
+read the OID and how; it does not prove *what they branch on*. Every
+"likely role" / "patching effect" claim above should be confirmed by
+running the patched dylib and observing behaviour.
diff --git a/research/hv_vmm_present_xref.json b/research/hv_vmm_present_xref.json
new file mode 100644
index 00000000..2d910fee
--- /dev/null
+++ b/research/hv_vmm_present_xref.json
@@ -0,0 +1,1351 @@
+[
+ {
+ "path": "Applications/CheckerBoard.app/CheckerBoard",
+ "consumers": [
+ {
+ "string_vma": "0x10006fb56",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x10006fb56",
+ "sect": "__text",
+ "xref_at": "0x1000192e8",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1000192d8 stur wzr, [x29, #-4]",
+ "0x1000192dc mov w8, #4",
+ "0x1000192e0 str x8, [sp]",
+ "0x1000192e4 adrp x0, #0x10006f000",
+ "0x1000192e8 add x0, x0, #0xb56",
+ "0x1000192ec sub x1, x29, #4",
+ "0x1000192f0 mov x2, sp",
+ "0x1000192f4 mov x3, #0",
+ "0x1000192f8 mov x4, #0",
+ "0x1000192fc bl #0x100048fa4"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "Applications/Family.app/Family",
+ "consumers": []
+ },
+ {
+ "path": "Applications/PeopleMessageService.app/PeopleMessageService",
+ "consumers": []
+ },
+ {
+ "path": "Applications/PeopleViewService.app/PeopleViewService",
+ "consumers": []
+ },
+ {
+ "path": "Applications/StoreKitUISceneService.app/StoreKitUISceneService",
+ "consumers": [
+ {
+ "string_vma": "0x1000a2ca0",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1000a2ca0",
+ "sect": "__text",
+ "xref_at": "0x100083f04",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x100083ef4 mov x1, #-1",
+ "0x100083ef8 bl #0x100090090",
+ "0x100083efc mov x21, x0",
+ "0x100083f00 adrp x0, #0x1000a2000",
+ "0x100083f04 add x0, x0, #0xca0",
+ "0x100083f08 mov x1, x19",
+ "0x100083f0c mov x2, x21",
+ "0x100083f10 mov x3, #0",
+ "0x100083f14 mov x4, #0",
+ "0x100083f18 bl #0x100090200"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/CoreServices/ClarityBoard.app/ClarityBoard",
+ "consumers": []
+ },
+ {
+ "path": "System/Library/DataClassMigrators/MobileActivationMigrator.migrator/MobileActivationMigrator",
+ "consumers": [
+ {
+ "string_vma": "0x72aa",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x72aa",
+ "sect": "__text",
+ "xref_at": "0x3024",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x3014 stur wzr, [x29, #-4]",
+ "0x3018 mov w8, #4",
+ "0x301c str x8, [sp]",
+ "0x3020 adrp x0, #0x7000",
+ "0x3024 add x0, x0, #0x2aa",
+ "0x3028 sub x1, x29, #4",
+ "0x302c mov x2, sp",
+ "0x3030 mov x3, #0",
+ "0x3034 mov x4, #0",
+ "0x3038 bl #0x3788"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/ExtensionKit/Extensions/HostInferenceProviderService.appex/HostInferenceProviderService",
+ "consumers": [
+ {
+ "string_vma": "0x1000112b0",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1000112b0",
+ "sect": "__text",
+ "xref_at": "0x10000b890",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x10000b880 str x8, [sp, #0x18]",
+ "0x10000b884 mov w8, #8",
+ "0x10000b888 stp x8, xzr, [sp, #8]",
+ "0x10000b88c adrp x0, #0x100011000",
+ "0x10000b890 add x0, x0, #0x2b0",
+ "0x10000b894 add x1, sp, #0x10",
+ "0x10000b898 add x2, sp, #8",
+ "0x10000b89c mov x3, #0",
+ "0x10000b8a0 mov x4, #0",
+ "0x10000b8a4 bl #0x1000103a4"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/Frameworks/CoreVideo.framework/CoreVideo",
+ "consumers": [
+ {
+ "string_vma": "0x19d17a4bb",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x19d17a4bb",
+ "sect": "__text",
+ "xref_at": "0x19d122eb8",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x19d122ea8 add x29, sp, #0x10",
+ "0x19d122eac mov w8, #4",
+ "0x19d122eb0 str x8, [sp, #8]",
+ "0x19d122eb4 adrp x0, #0x19d17a000",
+ "0x19d122eb8 add x0, x0, #0x4bb",
+ "0x19d122ebc adrp x1, #0x1ed568000",
+ "0x19d122ec0 add x1, x1, #0x5bc",
+ "0x19d122ec4 add x2, sp, #8",
+ "0x19d122ec8 mov x3, #0",
+ "0x19d122ecc mov x4, #0",
+ "0x19d122ed0 bl #0x19eaf7f30"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/Frameworks/ManagedAppDistribution.framework/Support/managedappdistributiond",
+ "consumers": [
+ {
+ "string_vma": "0x1006c25eb",
+ "section": "__TEXT,__cstring",
+ "class": "DEAD",
+ "xrefs": []
+ }
+ ]
+ },
+ {
+ "path": "System/Library/Frameworks/MediaToolbox.framework/MediaToolbox",
+ "consumers": [
+ {
+ "string_vma": "0x196ee0991",
+ "section": "__TEXT,__cstring",
+ "class": "DEAD",
+ "xrefs": []
+ }
+ ]
+ },
+ {
+ "path": "System/Library/Frameworks/SoundAnalysis.framework/SoundAnalysis",
+ "consumers": [
+ {
+ "string_vma": "0x1c9ad7cf0",
+ "section": "__TEXT,__cstring",
+ "class": "REFERENCED",
+ "xrefs": [
+ {
+ "string_vma": "0x1c9ad7cf0",
+ "sect": "__text",
+ "xref_at": "0x1c989ba18",
+ "class": "REFERENCED",
+ "snippet": [
+ "0x1c989ba08 stp x20, x19, [sp, #0x20]",
+ "0x1c989ba0c stp x29, x30, [sp, #0x30]",
+ "0x1c989ba10 add x29, sp, #0x30",
+ "0x1c989ba14 adrp x8, #0x1c9ad7000",
+ "0x1c989ba18 add x8, x8, #0xcf0",
+ "0x1c989ba1c sub x20, x8, #0x20",
+ "0x1c989ba20 mov w0, #4",
+ "0x1c989ba24 mov x1, #-1",
+ "0x1c989ba28 bl #0x1cca93270"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/Frameworks/StoreKit.framework/Support/storekitd",
+ "consumers": [
+ {
+ "string_vma": "0x10031a890",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x10031a890",
+ "sect": "__text",
+ "xref_at": "0x100176f6c",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x100176f5c mov x1, #-1",
+ "0x100176f60 bl #0x1002d586c",
+ "0x100176f64 mov x21, x0",
+ "0x100176f68 adrp x0, #0x10031a000",
+ "0x100176f6c add x0, x0, #0x890",
+ "0x100176f70 mov x1, x19",
+ "0x100176f74 mov x2, x21",
+ "0x100176f78 mov x3, #0",
+ "0x100176f7c mov x4, #0",
+ "0x100176f80 bl #0x1002d59ec"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/AAAFoundation.framework/AAAFoundation",
+ "consumers": [
+ {
+ "string_vma": "0x1c86581a1",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1c86581a1",
+ "sect": "__text",
+ "xref_at": "0x1c8650ee4",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1c8650ed4 stur wzr, [x29, #-4]",
+ "0x1c8650ed8 mov w8, #4",
+ "0x1c8650edc str x8, [sp]",
+ "0x1c8650ee0 adrp x0, #0x1c8658000",
+ "0x1c8650ee4 add x0, x0, #0x1a1",
+ "0x1c8650ee8 sub x1, x29, #4",
+ "0x1c8650eec mov x2, sp",
+ "0x1c8650ef0 mov x3, #0",
+ "0x1c8650ef4 mov x4, #0",
+ "0x1c8650ef8 bl #0x1cca79c30"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/AirPlaySupport.framework/AirPlaySupport",
+ "consumers": [
+ {
+ "string_vma": "0x2223618a0",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x2223618a0",
+ "sect": "__text",
+ "xref_at": "0x222306a28",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x222306a18 stur wzr, [x29, #-0x14]",
+ "0x222306a1c mov w8, #4",
+ "0x222306a20 str x8, [sp, #0x10]",
+ "0x222306a24 adrp x0, #0x222361000",
+ "0x222306a28 add x0, x0, #0x8a0",
+ "0x222306a2c sub x1, x29, #0x14",
+ "0x222306a30 add x2, sp, #0x10",
+ "0x222306a34 mov x3, #0",
+ "0x222306a38 mov x4, #0",
+ "0x222306a3c bl #0x223db21c0"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/AppStoreUtilities.framework/AppStoreUtilities",
+ "consumers": [
+ {
+ "string_vma": "0x23ff916d7",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x23ff916d7",
+ "sect": "__text",
+ "xref_at": "0x23ff8bfe4",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x23ff8bfd4 str x8, [sp, #0x18]",
+ "0x23ff8bfd8 mov w8, #4",
+ "0x23ff8bfdc str x8, [sp, #8]",
+ "0x23ff8bfe0 adrp x0, #0x23ff91000",
+ "0x23ff8bfe4 add x0, x0, #0x6d7",
+ "0x23ff8bfe8 adrp x1, #0x27e2bb000",
+ "0x23ff8bfec add x1, x1, #0x2b8",
+ "0x23ff8bff0 add x2, sp, #8",
+ "0x23ff8bff4 mov x3, #0",
+ "0x23ff8bff8 mov x4, #0",
+ "0x23ff8bffc bl #0x245f81640"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/ApplePushService.framework/ApplePushService",
+ "consumers": [
+ {
+ "string_vma": "0x1b235fd50",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1b235fd50",
+ "sect": "__text",
+ "xref_at": "0x1b2358d10",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1b2358d00 stur wzr, [x29, #-4]",
+ "0x1b2358d04 mov w8, #4",
+ "0x1b2358d08 str x8, [sp]",
+ "0x1b2358d0c adrp x0, #0x1b235f000",
+ "0x1b2358d10 add x0, x0, #0xd50",
+ "0x1b2358d14 sub x1, x29, #4",
+ "0x1b2358d18 mov x2, sp",
+ "0x1b2358d1c mov x3, #0",
+ "0x1b2358d20 mov x4, #0",
+ "0x1b2358d24 bl #0x1b2746fb0"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/ApplePushService.framework/apsd",
+ "consumers": [
+ {
+ "string_vma": "0x10015254a",
+ "section": "__TEXT,__cstring",
+ "class": "DEAD",
+ "xrefs": []
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/AppleNeuralEngine.framework/AppleNeuralEngine",
+ "consumers": [
+ {
+ "string_vma": "0x1ad297375",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1ad297375",
+ "sect": "__text",
+ "xref_at": "0x1ad248248",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1ad248238 stur wzr, [x29, #-4]",
+ "0x1ad24823c mov w8, #4",
+ "0x1ad248240 str x8, [sp]",
+ "0x1ad248244 adrp x0, #0x1ad297000",
+ "0x1ad248248 add x0, x0, #0x375",
+ "0x1ad24824c sub x1, x29, #4",
+ "0x1ad248250 mov x2, sp",
+ "0x1ad248254 mov x3, #0",
+ "0x1ad248258 mov x4, #0",
+ "0x1ad24825c bl #0x1b26f3a40"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/CoreALD.framework/CoreALD",
+ "consumers": []
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/CoreCDP.framework/CoreCDP",
+ "consumers": [
+ {
+ "string_vma": "0x1dedfa62d",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1dedfa62d",
+ "sect": "__text",
+ "xref_at": "0x1dedbb724",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1dedbb714 stur wzr, [x29, #-4]",
+ "0x1dedbb718 mov w8, #4",
+ "0x1dedbb71c str x8, [sp]",
+ "0x1dedbb720 adrp x0, #0x1dedfa000",
+ "0x1dedbb724 add x0, x0, #0x62d",
+ "0x1dedbb728 sub x1, x29, #4",
+ "0x1dedbb72c mov x2, sp",
+ "0x1dedbb730 mov x3, #0",
+ "0x1dedbb734 mov x4, #0",
+ "0x1dedbb738 bl #0x1e12ca690"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/CMCapture.framework/CMCapture",
+ "consumers": [
+ {
+ "string_vma": "0x1ad0e79d1",
+ "section": "__TEXT,__cstring",
+ "class": "DEAD",
+ "xrefs": []
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/AuthKit.framework/AuthKit",
+ "consumers": [
+ {
+ "string_vma": "0x1933c11de",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1933c11de",
+ "sect": "__text",
+ "xref_at": "0x19322f7e8",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x19322f7d8 mov x2, sp",
+ "0x19322f7dc mov x8, #4",
+ "0x19322f7e0 str x8, [sp]",
+ "0x19322f7e4 adrp x0, #0x1933c1000",
+ "0x19322f7e8 add x0, x0, #0x1de",
+ "0x19322f7ec mov x3, #0",
+ "0x19322f7f0 mov x4, #0",
+ "0x19322f7f4 bl #0x193b16710"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/AppStoreDaemon.framework/Support/appstored",
+ "consumers": [
+ {
+ "string_vma": "0x10046dae6",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x10046dae6",
+ "sect": "__text",
+ "xref_at": "0x1003e25b8",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1003e25a8 str x8, [sp, #0x18]",
+ "0x1003e25ac mov w8, #4",
+ "0x1003e25b0 str x8, [sp, #8]",
+ "0x1003e25b4 adrp x0, #0x10046d000",
+ "0x1003e25b8 add x0, x0, #0xae6",
+ "0x1003e25bc adrp x1, #0x1005aa000",
+ "0x1003e25c0 add x1, x1, #0xfd0",
+ "0x1003e25c4 add x2, sp, #8",
+ "0x1003e25c8 mov x3, #0",
+ "0x1003e25cc mov x4, #0",
+ "0x1003e25d0 bl #0x1004115c8"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/CorePrescription.framework/CorePrescription",
+ "consumers": [
+ {
+ "string_vma": "0x2471fe6ed",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x2471fe6ed",
+ "sect": "__text",
+ "xref_at": "0x2471ebc18",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x2471ebc08 stur wzr, [x29, #-4]",
+ "0x2471ebc0c mov w8, #4",
+ "0x2471ebc10 str x8, [sp]",
+ "0x2471ebc14 adrp x0, #0x2471fe000",
+ "0x2471ebc18 add x0, x0, #0x6ed",
+ "0x2471ebc1c sub x1, x29, #4",
+ "0x2471ebc20 mov x2, sp",
+ "0x2471ebc24 mov x3, #0",
+ "0x2471ebc28 mov x4, #0",
+ "0x2471ebc2c bl #0x24c021520"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/DataDetectorsUI.framework/PlugIns/com.apple.DataDetectorsUI.ActionsExtension.appex/com.apple.DataDetectorsUI.ActionsExtension",
+ "consumers": []
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/DeviceCheckInternal.framework/DeviceCheckInternal",
+ "consumers": [
+ {
+ "string_vma": "0x2487a018c",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x2487a018c",
+ "sect": "__text",
+ "xref_at": "0x24878d10c",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x24878d0fc stur wzr, [x29, #-4]",
+ "0x24878d100 mov w8, #4",
+ "0x24878d104 str x8, [sp]",
+ "0x24878d108 adrp x0, #0x2487a0000",
+ "0x24878d10c add x0, x0, #0x18c",
+ "0x24878d110 sub x1, x29, #4",
+ "0x24878d114 mov x2, sp",
+ "0x24878d118 mov x3, #0",
+ "0x24878d11c mov x4, #0",
+ "0x24878d120 bl #0x24c05d440"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/DVTInstrumentsUtilities.framework/DVTInstrumentsUtilities",
+ "consumers": [
+ {
+ "string_vma": "0x247f42ad0",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x247f42ad0",
+ "sect": "__text",
+ "xref_at": "0x247f1154c",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x247f1153c stur wzr, [x29, #-4]",
+ "0x247f11540 mov w8, #4",
+ "0x247f11544 str x8, [sp]",
+ "0x247f11548 adrp x0, #0x247f42000",
+ "0x247f1154c add x0, x0, #0xad0",
+ "0x247f11550 sub x1, x29, #4",
+ "0x247f11554 mov x2, sp",
+ "0x247f11558 mov x3, #0",
+ "0x247f1155c mov x4, #0",
+ "0x247f11560 bl #0x24c046270"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/DeviceIdentity.framework/DeviceIdentity",
+ "consumers": [
+ {
+ "string_vma": "0x2260a4b25",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x2260a4b25",
+ "sect": "__text",
+ "xref_at": "0x2260862a4",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x226086294 stur wzr, [x29, #-4]",
+ "0x226086298 mov w8, #4",
+ "0x22608629c str x8, [sp]",
+ "0x2260862a0 adrp x0, #0x2260a4000",
+ "0x2260862a4 add x0, x0, #0xb25",
+ "0x2260862a8 sub x1, x29, #4",
+ "0x2260862ac mov x2, sp",
+ "0x2260862b0 mov x3, #0",
+ "0x2260862b4 mov x4, #0",
+ "0x2260862b8 bl #0x22ac14380"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/CorePrescription.framework/XPCServices/CorePrescriptionService.xpc/CorePrescriptionService",
+ "consumers": [
+ {
+ "string_vma": "0x100092175",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x100092175",
+ "sect": "__text",
+ "xref_at": "0x100005064",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x100005054 stur wzr, [x29, #-4]",
+ "0x100005058 mov w8, #4",
+ "0x10000505c str x8, [sp]",
+ "0x100005060 adrp x0, #0x100092000",
+ "0x100005064 add x0, x0, #0x175",
+ "0x100005068 sub x1, x29, #4",
+ "0x10000506c mov x2, sp",
+ "0x100005070 mov x3, #0",
+ "0x100005074 mov x4, #0",
+ "0x100005078 bl #0x100082668"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/IDS.framework/identityservicesd.app/identityservicesd",
+ "consumers": []
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/IOSurfaceAccelerator.framework/IOSurfaceAccelerator",
+ "consumers": [
+ {
+ "string_vma": "0x2547cd71c",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x2547cd71c",
+ "sect": "__text",
+ "xref_at": "0x2547caa48",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x2547caa38 str wzr, [sp, #0xc]",
+ "0x2547caa3c mov w8, #4",
+ "0x2547caa40 str x8, [sp]",
+ "0x2547caa44 adrp x0, #0x2547cd000",
+ "0x2547caa48 add x0, x0, #0x71c",
+ "0x2547caa4c add x1, sp, #0xc",
+ "0x2547caa50 mov x2, sp",
+ "0x2547caa54 mov x3, #0",
+ "0x2547caa58 mov x4, #0",
+ "0x2547caa5c bl #0x259aa0060"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/EmailFoundation.framework/EmailFoundation",
+ "consumers": [
+ {
+ "string_vma": "0x1c61cd1f7",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1c61cd1f7",
+ "sect": "__text",
+ "xref_at": "0x1c6165e88",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1c6165e78 stur wzr, [x29, #-4]",
+ "0x1c6165e7c mov w8, #4",
+ "0x1c6165e80 str x8, [sp]",
+ "0x1c6165e84 adrp x0, #0x1c61cd000",
+ "0x1c6165e88 add x0, x0, #0x1f7",
+ "0x1c6165e8c sub x1, x29, #4",
+ "0x1c6165e90 mov x2, sp",
+ "0x1c6165e94 mov x3, #0",
+ "0x1c6165e98 mov x4, #0",
+ "0x1c6165e9c bl #0x1c6957f00"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/FindMyBase.framework/FindMyBase",
+ "consumers": [
+ {
+ "string_vma": "0x1df635420",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1df635420",
+ "sect": "__text",
+ "xref_at": "0x1df5fb8d4",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1df5fb8c4 stur wzr, [x29, #-0xc]",
+ "0x1df5fb8c8 mov w8, #4",
+ "0x1df5fb8cc str x8, [sp, #8]",
+ "0x1df5fb8d0 adrp x0, #0x1df635000",
+ "0x1df5fb8d4 add x0, x0, #0x420",
+ "0x1df5fb8d8 sub x1, x29, #0xc",
+ "0x1df5fb8dc add x2, sp, #8",
+ "0x1df5fb8e0 mov x3, #0",
+ "0x1df5fb8e4 mov x4, #0",
+ "0x1df5fb8e8 bl #0x1e12d9f10"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/IDSFoundation.framework/IDSFoundation",
+ "consumers": [
+ {
+ "string_vma": "0x1a7ea0fe4",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1a7ea0fe4",
+ "sect": "__text",
+ "xref_at": "0x1a7c65dc4",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1a7c65db4 str wzr, [sp, #0x1c]",
+ "0x1a7c65db8 mov w8, #4",
+ "0x1a7c65dbc str x8, [sp, #0x10]",
+ "0x1a7c65dc0 adrp x0, #0x1a7ea0000",
+ "0x1a7c65dc4 add x0, x0, #0xfe4",
+ "0x1a7c65dc8 add x1, sp, #0x1c",
+ "0x1a7c65dcc add x2, sp, #0x10",
+ "0x1a7c65dd0 mov x3, #0",
+ "0x1a7c65dd4 mov x4, #0",
+ "0x1a7c65dd8 bl #0x1ac565650"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/MagnifierSupport.framework/MagnifierSupport",
+ "consumers": [
+ {
+ "string_vma": "0x257d83670",
+ "section": "__TEXT,__cstring",
+ "class": "DEAD",
+ "xrefs": []
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/MobileActivation.framework/MobileActivation",
+ "consumers": [
+ {
+ "string_vma": "0x2598f06c6",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x2598f06c6",
+ "sect": "__text",
+ "xref_at": "0x2598e2618",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x2598e2608 stur wzr, [x29, #-4]",
+ "0x2598e260c mov w8, #4",
+ "0x2598e2610 str x8, [sp]",
+ "0x2598e2614 adrp x0, #0x2598f0000",
+ "0x2598e2618 add x0, x0, #0x6c6",
+ "0x2598e261c sub x1, x29, #4",
+ "0x2598e2620 mov x2, sp",
+ "0x2598e2624 mov x3, #0",
+ "0x2598e2628 mov x4, #0",
+ "0x2598e262c bl #0x259b2a0d0"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/PhotoFoundation.framework/PhotoFoundation",
+ "consumers": [
+ {
+ "string_vma": "0x1d8ba907c",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1d8ba907c",
+ "sect": "__text",
+ "xref_at": "0x1d8b9e7fc",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1d8b9e7ec stur wzr, [x29, #-4]",
+ "0x1d8b9e7f0 mov w8, #4",
+ "0x1d8b9e7f4 str x8, [sp]",
+ "0x1d8b9e7f8 adrp x0, #0x1d8ba9000",
+ "0x1d8b9e7fc add x0, x0, #0x7c",
+ "0x1d8b9e800 sub x1, x29, #4",
+ "0x1d8b9e804 mov x2, sp",
+ "0x1d8b9e808 mov x3, #0",
+ "0x1d8b9e80c mov x4, #0",
+ "0x1d8b9e810 bl #0x1da721db0"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/Recon3D.framework/Recon3D",
+ "consumers": [
+ {
+ "string_vma": "0x1cc912676",
+ "section": "__TEXT,__cstring",
+ "class": "DEAD",
+ "xrefs": []
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/NeuralNetworks.framework/NeuralNetworks",
+ "consumers": [
+ {
+ "string_vma": "0x25bb594c0",
+ "section": "__TEXT,__cstring",
+ "class": "DEAD",
+ "xrefs": []
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/SonicKit.framework/SonicKit",
+ "consumers": []
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/VideosUI.framework/VideosUI",
+ "consumers": []
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/VFX.framework/VFX",
+ "consumers": [
+ {
+ "string_vma": "0x1aff31940",
+ "section": "__TEXT,__cstring",
+ "class": "DEAD",
+ "xrefs": []
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/RenderBox.framework/RenderBox",
+ "consumers": [
+ {
+ "string_vma": "0x195e54555",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x195e54555",
+ "sect": "__text",
+ "xref_at": "0x195d344a0",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x195d34490 stur wzr, [x29, #-4]",
+ "0x195d34494 mov w8, #4",
+ "0x195d34498 str x8, [sp]",
+ "0x195d3449c adrp x0, #0x195e54000",
+ "0x195d344a0 add x0, x0, #0x555",
+ "0x195d344a4 sub x1, x29, #4",
+ "0x195d344a8 mov x2, sp",
+ "0x195d344ac mov x3, #0",
+ "0x195d344b0 mov x4, #0",
+ "0x195d344b4 bl #0x19a8c1300"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/TrialServer.framework/TrialServer",
+ "consumers": [
+ {
+ "string_vma": "0x26f32471e",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x26f32471e",
+ "sect": "__text",
+ "xref_at": "0x26f224120",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x26f224110 str wzr, [sp, #0xc]",
+ "0x26f224114 mov w8, #4",
+ "0x26f224118 str x8, [sp]",
+ "0x26f22411c adrp x0, #0x26f324000",
+ "0x26f224120 add x0, x0, #0x71e",
+ "0x26f224124 add x1, sp, #0xc",
+ "0x26f224128 mov x2, sp",
+ "0x26f22412c mov x3, #0",
+ "0x26f224130 mov x4, #0",
+ "0x26f224134 bl #0x274313ed0"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/VisionKitCore.framework/VisionKitCore",
+ "consumers": [
+ {
+ "string_vma": "0x1b442d9cf",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1b442d9cf",
+ "sect": "__text",
+ "xref_at": "0x1b435f0a4",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1b435f094 stur wzr, [x29, #-4]",
+ "0x1b435f098 mov w8, #4",
+ "0x1b435f09c str x8, [sp]",
+ "0x1b435f0a0 adrp x0, #0x1b442d000",
+ "0x1b435f0a4 add x0, x0, #0x9cf",
+ "0x1b435f0a8 sub x1, x29, #4",
+ "0x1b435f0ac mov x2, sp",
+ "0x1b435f0b0 mov x3, #0",
+ "0x1b435f0b4 mov x4, #0",
+ "0x1b435f0b8 bl #0x1b8c77210"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/WatchdogServiceManagement.framework/WatchdogServiceManagement",
+ "consumers": [
+ {
+ "string_vma": "0x272725b23",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x272725b23",
+ "sect": "__text",
+ "xref_at": "0x272725114",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x272725104 stur wzr, [x29, #-4]",
+ "0x272725108 mov w8, #4",
+ "0x27272510c str x8, [sp]",
+ "0x272725110 adrp x0, #0x272725000",
+ "0x272725114 add x0, x0, #0xb23",
+ "0x272725118 sub x1, x29, #4",
+ "0x27272511c mov x2, sp",
+ "0x272725120 mov x3, #0",
+ "0x272725124 mov x4, #0",
+ "0x272725128 bl #0x274352f80"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "usr/sbin/absd",
+ "consumers": []
+ },
+ {
+ "path": "usr/lib/libMobileGestalt.dylib",
+ "consumers": [
+ {
+ "string_vma": "0x1b0201c60",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1b0201c60",
+ "sect": "__text",
+ "xref_at": "0x1b0198d00",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1b0198cf0 stur wzr, [x29, #-4]",
+ "0x1b0198cf4 mov w8, #4",
+ "0x1b0198cf8 str x8, [sp]",
+ "0x1b0198cfc adrp x0, #0x1b0201000",
+ "0x1b0198d00 add x0, x0, #0xc60",
+ "0x1b0198d04 sub x1, x29, #4",
+ "0x1b0198d08 mov x2, sp",
+ "0x1b0198d0c mov x3, #0",
+ "0x1b0198d10 mov x4, #0",
+ "0x1b0198d14 bl #0x1b27216f0"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "usr/sbin/fairplayd.H2",
+ "consumers": []
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/caulk.framework/caulk",
+ "consumers": [
+ {
+ "string_vma": "0x27512204f",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x27512204f",
+ "sect": "__text",
+ "xref_at": "0x275112808",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x2751127f8 stur wzr, [x29, #-4]",
+ "0x2751127fc mov w8, #4",
+ "0x275112800 str x8, [sp]",
+ "0x275112804 adrp x0, #0x275122000",
+ "0x275112808 add x0, x0, #0x4f",
+ "0x27511280c sub x1, x29, #4",
+ "0x275112810 mov x2, sp",
+ "0x275112814 mov x3, #0",
+ "0x275112818 mov x4, #0",
+ "0x27511281c bl #0x277e067e0"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "usr/sbin/bluetoothd",
+ "consumers": [
+ {
+ "string_vma": "0x1009c2a7f",
+ "section": "__TEXT,__cstring",
+ "class": "DEAD",
+ "xrefs": []
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/WebGPU.framework/WebGPU",
+ "consumers": [
+ {
+ "string_vma": "0x225722b6c",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x225722b6c",
+ "sect": "__text",
+ "xref_at": "0x22552c2bc",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x22552c2ac str wzr, [sp, #0xc]",
+ "0x22552c2b0 mov w8, #4",
+ "0x22552c2b4 str x8, [sp]",
+ "0x22552c2b8 adrp x0, #0x225722000",
+ "0x22552c2bc add x0, x0, #0xb6c",
+ "0x22552c2c0 add x1, sp, #0xc",
+ "0x22552c2c4 mov x2, sp",
+ "0x22552c2c8 mov x3, #0",
+ "0x22552c2cc mov x4, #0",
+ "0x22552c2d0 bl #0x22ac033e0"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/Frameworks/CoreML.framework/CoreML",
+ "consumers": [
+ {
+ "string_vma": "0x1950ed0d0",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1950ed0d0",
+ "xref_at": "0x194c3c754",
+ "sect": "__text",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x194c3c744 stur wzr, [x29, #-0xc]",
+ "0x194c3c748 mov w8, #4",
+ "0x194c3c74c str x8, [sp, #8]",
+ "0x194c3c750 adrp x0, #0x1950ed000",
+ "0x194c3c754 add x0, x0, #0xd0",
+ "0x194c3c758 sub x1, x29, #0xc",
+ "0x194c3c75c add x2, sp, #8",
+ "0x194c3c760 mov x3, #0",
+ "0x194c3c764 mov x4, #0",
+ "0x194c3c768 bl #0x19a8b2010",
+ "0x194c3c76c cmp w0, #0",
+ "0x194c3c770 ldur w8, [x29, #-0xc]",
+ "0x194c3c774 ccmp w8, #0, #4, eq",
+ "0x194c3c778 cset w8, ne",
+ "0x194c3c77c adrp x9, #0x1eaec5000",
+ "0x194c3c780 strb w8, [x9, #0x198]",
+ "0x194c3c784 ldur x8, [x29, #-8]",
+ "0x194c3c788 adrp x9, #0x1e69e9000",
+ "0x194c3c78c ldr x9, [x9, #0x840]",
+ "0x194c3c790 ldr x9, [x9]",
+ "0x194c3c794 cmp x9, x8",
+ "0x194c3c798 b.ne #0x194c3c7a8",
+ "0x194c3c79c ldp x29, x30, [sp, #0x20]",
+ "0x194c3c7a0 add sp, sp, #0x30",
+ "0x194c3c7a4 retab ",
+ "0x194c3c7a8 bl #0x19a8b0300",
+ "0x194c3c7ac pacibsp ",
+ "0x194c3c7b0 stp x20, x19, [sp, #-0x20]!"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/CoreRE.framework/CoreRE",
+ "consumers": [
+ {
+ "string_vma": "0x1e3171b20",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1e3171b20",
+ "xref_at": "0x1e2790f28",
+ "sect": "__text",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1e2790f18 stur wzr, [x29, #-4]",
+ "0x1e2790f1c mov w8, #4",
+ "0x1e2790f20 str x8, [sp]",
+ "0x1e2790f24 adrp x0, #0x1e3171000",
+ "0x1e2790f28 add x0, x0, #0xb20",
+ "0x1e2790f2c sub x1, x29, #4",
+ "0x1e2790f30 mov x2, sp",
+ "0x1e2790f34 mov x3, #0",
+ "0x1e2790f38 mov x4, #0",
+ "0x1e2790f3c bl #0x1e6908200",
+ "0x1e2790f40 cbnz w0, #0x1e2790f58",
+ "0x1e2790f44 ldur w8, [x29, #-4]",
+ "0x1e2790f48 cmp w8, #0",
+ "0x1e2790f4c cset w8, eq",
+ "0x1e2790f50 adrp x9, #0x1ee186000",
+ "0x1e2790f54 strb w8, [x9, #0xc8]",
+ "0x1e2790f58 ldp x29, x30, [sp, #0x10]",
+ "0x1e2790f5c add sp, sp, #0x20",
+ "0x1e2790f60 retab ",
+ "0x1e2790f64 pacibsp ",
+ "0x1e2790f68 stp x20, x19, [sp, #-0x20]!",
+ "0x1e2790f6c stp x29, x30, [sp, #0x10]",
+ "0x1e2790f70 add x29, sp, #0x10",
+ "0x1e2790f74 adrp x8, #0x1e6974000",
+ "0x1e2790f78 ldr x0, [x8]",
+ "0x1e2790f7c bl #0x1e323d4c0",
+ "0x1e2790f80 bl #0x1e69076d0",
+ "0x1e2790f84 mov x19, x0"
+ ]
+ }
+ ]
+ }
+ ]
+ },
+ {
+ "path": "System/Library/PrivateFrameworks/Espresso.framework/Espresso",
+ "consumers": [
+ {
+ "string_vma": "0x1949204e5",
+ "section": "__TEXT,__cstring",
+ "class": "CANONICAL",
+ "xrefs": [
+ {
+ "string_vma": "0x1949204e5",
+ "xref_at": "0x193bef83c",
+ "sect": "__text",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x193bef82c stur wzr, [x29, #-4]",
+ "0x193bef830 mov w8, #4",
+ "0x193bef834 str x8, [sp]",
+ "0x193bef838 adrp x0, #0x194920000",
+ "0x193bef83c add x0, x0, #0x4e5",
+ "0x193bef840 sub x1, x29, #4",
+ "0x193bef844 mov x2, sp",
+ "0x193bef848 mov x3, #0",
+ "0x193bef84c mov x4, #0",
+ "0x193bef850 bl #0x19a8ac6e0",
+ "0x193bef854 cbz w0, #0x193bef864",
+ "0x193bef858 ldp x29, x30, [sp, #0x10]",
+ "0x193bef85c add sp, sp, #0x20",
+ "0x193bef860 retab ",
+ "0x193bef864 ldur w8, [x29, #-4]",
+ "0x193bef868 cmp w8, #0",
+ "0x193bef86c cset w8, ne",
+ "0x193bef870 adrp x9, #0x1eaeaf000",
+ "0x193bef874 strb w8, [x9, #0x4f8]",
+ "0x193bef878 ldp x29, x30, [sp, #0x10]",
+ "0x193bef87c add sp, sp, #0x20",
+ "0x193bef880 retab ",
+ "0x193bef884 pacibsp ",
+ "0x193bef888 sub sp, sp, #0x20",
+ "0x193bef88c stp x29, x30, [sp, #0x10]",
+ "0x193bef890 add x29, sp, #0x10",
+ "0x193bef894 add x8, sp, #8",
+ "0x193bef898 mov x1, #0"
+ ]
+ },
+ {
+ "string_vma": "0x1949204e5",
+ "xref_at": "0x194236a24",
+ "sect": "__text",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x194236a14 stur wzr, [x29, #-4]",
+ "0x194236a18 mov w8, #4",
+ "0x194236a1c str x8, [sp]",
+ "0x194236a20 adrp x0, #0x194920000",
+ "0x194236a24 add x0, x0, #0x4e5",
+ "0x194236a28 sub x1, x29, #4",
+ "0x194236a2c mov x2, sp",
+ "0x194236a30 mov x3, #0",
+ "0x194236a34 mov x4, #0",
+ "0x194236a38 bl #0x19a8ac6e0",
+ "0x194236a3c cbnz w0, #0x194236a4c",
+ "0x194236a40 ldur w8, [x29, #-4]",
+ "0x194236a44 adrp x9, #0x1ed6d6000",
+ "0x194236a48 str w8, [x9, #0x350]",
+ "0x194236a4c ldp x29, x30, [sp, #0x10]",
+ "0x194236a50 add sp, sp, #0x20",
+ "0x194236a54 retab ",
+ "0x194236a58 pacibsp ",
+ "0x194236a5c stp x29, x30, [sp, #-0x10]!",
+ "0x194236a60 mov x29, sp",
+ "0x194236a64 adrp x8, #0x1e698c000",
+ "0x194236a68 ldr x0, [x8, #0xd58]",
+ "0x194236a6c bl #0x1949b2400",
+ "0x194236a70 adrp x8, #0x1ed6d6000",
+ "0x194236a74 strb w0, [x8, #0x368]",
+ "0x194236a78 ldp x29, x30, [sp], #0x10",
+ "0x194236a7c retab ",
+ "0x194236a80 adrp x8, #0x1ed6d6000"
+ ]
+ },
+ {
+ "string_vma": "0x1949204e5",
+ "xref_at": "0x1942873cc",
+ "sect": "__text",
+ "class": "CANONICAL",
+ "snippet": [
+ "0x1942873bc stur wzr, [x29, #-4]",
+ "0x1942873c0 mov w8, #4",
+ "0x1942873c4 str x8, [sp]",
+ "0x1942873c8 adrp x0, #0x194920000",
+ "0x1942873cc add x0, x0, #0x4e5",
+ "0x1942873d0 sub x1, x29, #4",
+ "0x1942873d4 mov x2, sp",
+ "0x1942873d8 mov x3, #0",
+ "0x1942873dc mov x4, #0",
+ "0x1942873e0 bl #0x19a8ac6e0",
+ "0x1942873e4 cbnz w0, #0x1942873f4",
+ "0x1942873e8 ldur w8, [x29, #-4]",
+ "0x1942873ec adrp x9, #0x1ed6d6000",
+ "0x1942873f0 str w8, [x9, #0x1d8]",
+ "0x1942873f4 ldp x29, x30, [sp, #0x10]",
+ "0x1942873f8 add sp, sp, #0x20",
+ "0x1942873fc retab ",
+ "0x194287400 b #0x19a8aad90",
+ "0x194287404 b #0x19a8aad90",
+ "0x194287408 b #0x19a8aad90",
+ "0x19428740c b #0x19a8aad90",
+ "0x194287410 b #0x19a8aad90",
+ "0x194287414 pacibsp ",
+ "0x194287418 sub sp, sp, #0x60",
+ "0x19428741c stp x22, x21, [sp, #0x30]",
+ "0x194287420 stp x20, x19, [sp, #0x40]",
+ "0x194287424 stp x29, x30, [sp, #0x50]",
+ "0x194287428 add x29, sp, #0x50"
+ ]
+ }
+ ]
+ }
+ ]
+ }
+]
\ No newline at end of file
diff --git a/scripts/cfw_install_exp.sh b/scripts/cfw_install_exp.sh
old mode 100755
new mode 100644
index 4db72bf2..4c683dd0
--- a/scripts/cfw_install_exp.sh
+++ b/scripts/cfw_install_exp.sh
@@ -2,11 +2,24 @@
# cfw_install_exp.sh — Install base CFW + JB extensions + EXP experimental
# patches on vphone via SSH ramdisk.
#
-# Currently runs the same set of phases as the JB install script (JB-1..JB-5):
-# launchd jetsam patch, dylib injection, procursus bootstrap, BaseBin hook
-# deployment, first-boot setup. Subsequent commits in this branch add the
-# EXP-only experimental phases (DSC pre-step, EXP-JB-3.5 watchdogd, EXP-JB-6
-# post-restore DT, EXP-JB-7 build-version) on top.
+# Runs the base CFW installer first (phases 1-7), the JB-specific
+# modifications (launchd jetsam patch, dylib injection, procursus bootstrap,
+# BaseBin hook deployment, JB-1..JB-5), and additionally the experimental
+# phases — labeled `EXP-JB-N` to make their EXP-only scope obvious:
+# - Pre-step : byte-5 mangle of kern.hv_vmm_present in DSC dylibs
+# (paired with KernelEXPPatcher's kernel-side rename).
+# - EXP-JB-3.5 : surgical 2-insn patch of watchdogd's hv_vmm cache +
+# slot re-attest of the standalone Mach-O.
+# - EXP-JB-6 : post-restore DT identity rewrite (root model /
+# target-type / compatible[0]) inside
+# /mnt5/.../devicetree.img4.
+# - EXP-JB-7 : optional ProductBuildVersion rewrite in
+# SystemVersion.plist (gated on the SPOOF_BUILD env var).
+#
+# Stages JB-1..JB-5 remain genuine jailbreak phases (inherited from the JB
+# pipeline). Stages prefixed EXP-JB-* are EXP-exclusive — they do NOT run
+# when the JB or DEV install scripts are used. JB and DEV remain on their
+# pre-experimental baseline.
#
# Prerequisites (in addition to cfw_install.sh requirements):
# - cfw_jb_input/ or resources/cfw_jb_input.tar.zst present
@@ -38,6 +51,64 @@ PYTHON3="$(_resolve_python3)"
# ════════════════════════════════════════════════════════════════
echo "[*] cfw_install_exp.sh — Installing CFW + JB extensions + EXP experimental patches..."
echo ""
+
+# ────────────────────────────────────────────────────────────────────
+# Pre-step: patch hv_vmm_present user-mode consumers in the SystemOS
+# Cryptex's DSC chunks BEFORE running cfw_install.sh.
+#
+# cfw_install.sh decrypts the SystemOS Cryptex AEA into
+# $TEMP_DIR/CryptexSystemOS.dmg and reuses it on subsequent runs. We
+# pre-create that decrypted DMG here, mount it, patch the DSC chunks
+# in place, unmount, and let cfw_install.sh pick up the cached
+# (already-patched) DMG. cfw_install.sh itself is unmodified — this
+# is the EXP variant's device-like user-mode patching, kept out of
+# the JB install path so JB remains unaffected.
+# ────────────────────────────────────────────────────────────────────
+VM_DIR_ABS="$(cd "${VM_DIR:-.}" && pwd)"
+JB_TEMP_DIR="$VM_DIR_ABS/.cfw_temp"
+JB_SYSOS_DMG="$JB_TEMP_DIR/CryptexSystemOS.dmg"
+JB_MNT_SYSOS="$JB_TEMP_DIR/mnt_sysos_hv_vmm"
+mkdir -p "$JB_TEMP_DIR"
+
+# Find the restore directory (same logic as cfw_install.sh)
+JB_RESTORE_DIR=""
+for d in "$VM_DIR_ABS"/iPhone*_Restore; do
+ [[ -d "$d" ]] && { JB_RESTORE_DIR="$d"; break; }
+done
+
+if [[ -z "$JB_RESTORE_DIR" ]]; then
+ echo "[!] hv_vmm DSC patch: no restore directory found, skipping"
+elif [[ ! -f "$JB_SYSOS_DMG" ]]; then
+ # Not yet decrypted — decrypt to the cache location cfw_install.sh expects.
+ echo "[*] hv_vmm DSC patch: decrypting SystemOS into cache..."
+ JB_CRYPTEX_SYSOS=$("$PYTHON3" "$SCRIPT_DIR/patchers/cfw.py" cryptex-paths "$JB_RESTORE_DIR/iPhone-BuildManifest.plist" | head -1)
+ JB_AEA_KEY=$(ipsw fw aea --key "$JB_RESTORE_DIR/$JB_CRYPTEX_SYSOS")
+ aea decrypt -i "$JB_RESTORE_DIR/$JB_CRYPTEX_SYSOS" -o "$JB_SYSOS_DMG" -key-value "$JB_AEA_KEY"
+fi
+
+if [[ -f "$JB_SYSOS_DMG" ]]; then
+ # Mount, patch chunks, unmount. Idempotent: re-running on an already
+ # patched DMG is a no-op (patcher detects already-patched form).
+ echo "[*] hv_vmm DSC patch: mounting cached SystemOS DMG..."
+ mkdir -p "$JB_MNT_SYSOS"
+ sudo hdiutil detach "$JB_MNT_SYSOS" -force 2>/dev/null || true
+ sudo hdiutil attach -mountpoint "$JB_MNT_SYSOS" "$JB_SYSOS_DMG" -nobrowse -owners off
+
+ JB_DSC_CHUNKS_DIR="$JB_MNT_SYSOS/System/Library/Caches/com.apple.dyld"
+ if [[ -d "$JB_DSC_CHUNKS_DIR" ]]; then
+ echo "[*] hv_vmm DSC patch: patching chunks under $JB_DSC_CHUNKS_DIR..."
+ "$SCRIPT_DIR/patch_hv_vmm_userland.sh" dsc "$JB_DSC_CHUNKS_DIR"
+ echo "[+] hv_vmm DSC patch: chunks patched"
+ else
+ echo "[-] hv_vmm DSC patch: $JB_DSC_CHUNKS_DIR not found, skipping"
+ fi
+
+ sudo hdiutil detach "$JB_MNT_SYSOS" -force
+fi
+
+# Now run the regular CFW install. It will see the cached (patched)
+# CryptexSystemOS.dmg and use it as-is, so the patched DSC chunks land
+# in /mnt1/System/Cryptexes/OS on the device.
CFW_SKIP_HALT=1 zsh "$SCRIPT_DIR/cfw_install.sh" "$VM_DIR"
# ════════════════════════════════════════════════════════════════
diff --git a/scripts/patch_hv_vmm_userland.sh b/scripts/patch_hv_vmm_userland.sh
new file mode 100755
index 00000000..f996e4c9
--- /dev/null
+++ b/scripts/patch_hv_vmm_userland.sh
@@ -0,0 +1,66 @@
+#!/bin/zsh
+# patch_hv_vmm_userland.sh — Apply the user-mode hv_vmm_present patch.
+#
+# Two operations, chosen by the first arg:
+#
+# dsc <chunks_dir>
+# Patch the canonical sysctlbyname("kern.hv_vmm_present", ...) sites
+# inside the DSC chunks at <chunks_dir>. <chunks_dir> is the directory
+# holding `dyld_shared_cache_arm64e[.NN]` files, typically the
+# mounted SystemOS Cryptex's `System/Library/Caches/com.apple.dyld/`.
+# Skips the compute/accel dylibs (CoreML, Espresso, ANE, CoreRE,
+# RenderBox, WebGPU, caulk, IOSurfaceAccelerator).
+#
+# standalone <binary>
+# Patch a single standalone Mach-O file in place. Idempotent.
+# Caller is responsible for re-signing (ldid).
+#
+# This script is a thin wrapper around `scripts/patchers/cfw.py`. It
+# exists so cfw_install_dev.sh and cfw_install_jb.sh can call a single
+# entry point without duplicating Python venv/python3 resolution logic.
+#
+# Used by: cfw_install_dev.sh, cfw_install_jb.sh
+
+set -euo pipefail
+
+SCRIPT_DIR="${0:a:h}"
+
+[[ -n "${_VPHONE_PATH:-}" ]] && export PATH="$_VPHONE_PATH"
+
+_resolve_python3() {
+ local venv_py="${SCRIPT_DIR:h}/.venv/bin/python3"
+ if [[ -x "$venv_py" ]]; then
+ echo "$venv_py"
+ else
+ command -v python3 || true
+ fi
+}
+PYTHON3="$(_resolve_python3)"
+
+usage() {
+ cat <<EOF >&2
+Usage:
+ $0 dsc <chunks_dir>
+ $0 standalone <binary>
+EOF
+ exit 2
+}
+
+(( $# >= 1 )) || usage
+op="$1"; shift
+
+case "$op" in
+ dsc)
+ (( $# >= 1 )) || usage
+ echo "[*] Patching hv_vmm_present consumers in DSC chunks under: $1"
+ "$PYTHON3" "$SCRIPT_DIR/patchers/cfw.py" patch-hv-vmm-dsc "$1"
+ ;;
+ standalone)
+ (( $# >= 1 )) || usage
+ echo "[*] Patching hv_vmm_present consumers in: $1"
+ "$PYTHON3" "$SCRIPT_DIR/patchers/cfw.py" patch-hv-vmm "$1"
+ ;;
+ *)
+ usage
+ ;;
+esac
diff --git a/scripts/patchers/cfw.py b/scripts/patchers/cfw.py
index 116096ba..3aff4db4 100755
--- a/scripts/patchers/cfw.py
+++ b/scripts/patchers/cfw.py
@@ -23,6 +23,12 @@ Commands:
patch-launchd-jetsam <binary>
Patch launchd jetsam panic guard to avoid initproc crash loop.
+ patch-hv-vmm-dsc <chunks_dir> [--dry-run]
+ Same patch, applied in place to the DSC chunks under
+ <chunks_dir> (e.g. /System/Library/Caches/com.apple.dyld inside
+ the mounted SystemOS Cryptex). Targets a fixed list of identity,
+ store, and consumer-service dylibs; skips compute/accel libs.
+
inject-daemons <launchd.plist> <daemon_dir>
Inject bash/dropbear/trollvnc into launchd.plist.
@@ -32,6 +38,7 @@ Commands:
Dependencies:
pip install capstone keystone-engine
+ ipsw CLI in $PATH (only required for patch-hv-vmm-dsc, experimental variant only)
"""
import os
@@ -46,12 +53,14 @@ if __name__ == "__main__":
from patchers.cfw_patch_cache_loader import patch_launchd_cache_loader
from patchers.cfw_patch_mobileactivationd import patch_mobileactivationd
from patchers.cfw_patch_jetsam import patch_launchd_jetsam
+ from patchers.cfw_patch_hv_vmm_dsc import patch_hv_vmm_in_dsc
from patchers.cfw_daemons import parse_cryptex_paths, inject_daemons
else:
from .cfw_patch_seputil import patch_seputil
from .cfw_patch_cache_loader import patch_launchd_cache_loader
from .cfw_patch_mobileactivationd import patch_mobileactivationd
from .cfw_patch_jetsam import patch_launchd_jetsam
+ from .cfw_patch_hv_vmm_dsc import patch_hv_vmm_in_dsc
from .cfw_daemons import parse_cryptex_paths, inject_daemons
@@ -98,6 +107,14 @@ def main():
if not patch_launchd_jetsam(sys.argv[2]):
sys.exit(1)
+ elif cmd == "patch-hv-vmm-dsc":
+ if len(sys.argv) < 3:
+ print("Usage: patch_cfw.py patch-hv-vmm-dsc <chunks_dir> [--dry-run]")
+ sys.exit(1)
+ dry_run = "--dry-run" in sys.argv[3:]
+ results = patch_hv_vmm_in_dsc(sys.argv[2], dry_run=dry_run)
+ sys.exit(0)
+
elif cmd == "inject-daemons":
if len(sys.argv) < 4:
print("Usage: patch_cfw.py inject-daemons <launchd.plist> <daemon_dir>")
@@ -129,7 +146,7 @@ def main():
print(f"Unknown command: {cmd}")
print("Commands: cryptex-paths, patch-seputil, patch-launchd-cache-loader,")
print(" patch-mobileactivationd, patch-launchd-jetsam,")
- print(" inject-daemons, inject-dylib")
+ print(" patch-hv-vmm-dsc, inject-daemons, inject-dylib")
sys.exit(1)
diff --git a/scripts/patchers/cfw_dsc_chunks.py b/scripts/patchers/cfw_dsc_chunks.py
new file mode 100644
index 00000000..8ba10547
--- /dev/null
+++ b/scripts/patchers/cfw_dsc_chunks.py
@@ -0,0 +1,361 @@
+"""dyld shared cache (split / chunked) byte-level helper.
+
+The iOS 26 SystemOS Cryptex ships the DSC as multiple files:
+
+ /System/Library/Caches/com.apple.dyld/dyld_shared_cache_arm64e
+ /System/Library/Caches/com.apple.dyld/dyld_shared_cache_arm64e.01
+ /System/Library/Caches/com.apple.dyld/dyld_shared_cache_arm64e.02
+ ...
+
+The first file (no suffix) holds the cache header; suffixed files
+(`.NN`) hold subcache regions. Together they map a contiguous virtual
+address space.
+
+This module gives us:
+
+ chunks = DSCChunks(chunks_dir)
+ chunks.bytes_at_vma(vma, length) -> bytes
+ chunks.find_chunk_for_vma(vma) -> (chunk_path, file_offset)
+ chunks.write_at_vma(vma, data) -> None # persists to chunk on disk
+
+It does *not* extract dylibs. It treats the DSC as a flat virtual
+address space and gives byte-level access. That's the minimum we need
+to apply 4-byte instruction patches at known vmaddrs.
+
+The dyld_cache_header layout we use is intentionally minimal:
+ char magic[16];
+ uint32_t mappingOffset;
+ uint32_t mappingCount;
+ ...
+ uint64_t sharedRegionStart; (offset 0x60 in modern caches)
+ uint64_t sharedRegionSize;
+ ...
+
+The dyld_cache_mapping_info entries (mappingCount of them) each look:
+ uint64_t address;
+ uint64_t size;
+ uint64_t fileOffset;
+ uint32_t maxProt;
+ uint32_t initProt;
+
+For chunked caches (.NN files), each chunk has its own header with its
+own mappings; the mappings address ranges are non-overlapping across
+chunks.
+
+Reference: dyld project, dyld_cache_format.h.
+"""
+
+import os
+import struct
+import re
+import glob
+
+
+CHUNK_GLOB_NAMES = (
+ "dyld_shared_cache_arm64e",
+ "dyld_shared_cache_arm64e.[0-9]",
+ "dyld_shared_cache_arm64e.[0-9][0-9]",
+ "dyld_shared_cache_arm64e.symbols",
+)
+
+
+def _enumerate_chunks(chunks_dir):
+ """Return list of chunk file paths in numeric order; the unsuffixed file
+ first, then .01, .02, ... `.symbols` is excluded (it doesn't map code).
+ """
+ names = []
+ for name in os.listdir(chunks_dir):
+ if not name.startswith("dyld_shared_cache_arm64e"):
+ continue
+ if name.endswith(".symbols") or name.endswith(".map"):
+ continue
+ names.append(name)
+
+ def sort_key(n):
+ # base file (no suffix) first, then .01, .02, ...
+ m = re.match(r"^dyld_shared_cache_arm64e(?:\.(\d+))?$", n)
+ if m:
+ return (0, int(m.group(1)) if m.group(1) else -1)
+ return (1, n)
+
+ names.sort(key=sort_key)
+ return [os.path.join(chunks_dir, n) for n in names]
+
+
+def _parse_chunk_mappings(path):
+ """Return list of dicts {addr, size, file_off, max_prot, init_prot}
+ for one chunk file.
+
+ Only the standard dyld_cache_mapping_info entries are used. Slide-info
+ mappings (`mappingWithSlideOffset`) are ignored — for our purposes we
+ only care about which virtual address comes from which file offset,
+ and the standard mappings cover that.
+
+ `init_prot` is the VM_PROT mask: VM_PROT_READ=1, VM_PROT_WRITE=2,
+ VM_PROT_EXECUTE=4.
+ """
+ with open(path, "rb") as f:
+ head = f.read(0x100)
+ magic = head[:16].rstrip(b"\x00")
+ # Modern arm64e caches start with "dyld_v1 arm64e"
+ if not magic.startswith(b"dyld"):
+ return []
+
+ mapping_off = struct.unpack_from("<I", head, 16)[0]
+ mapping_cnt = struct.unpack_from("<I", head, 20)[0]
+ if mapping_cnt > 64 or mapping_off > 0x10000:
+ # sanity bounds
+ return []
+
+ with open(path, "rb") as f:
+ f.seek(mapping_off)
+ raw = f.read(mapping_cnt * 32)
+
+ out = []
+ for i in range(mapping_cnt):
+ off = i * 32
+ addr, size, file_off, max_prot, init_prot = \
+ struct.unpack_from("<QQQII", raw, off)
+ out.append({
+ "addr": addr, "size": size, "file_off": file_off,
+ "max_prot": max_prot, "init_prot": init_prot,
+ })
+ return out
+
+
+class DSCChunks:
+ """Lazy random-access byte view over a chunked dyld shared cache."""
+
+ def __init__(self, chunks_dir):
+ self.chunks_dir = chunks_dir
+ self._chunk_paths = _enumerate_chunks(chunks_dir)
+ if not self._chunk_paths:
+ raise FileNotFoundError(
+ f"No dyld_shared_cache_arm64e* chunks found under {chunks_dir!r}"
+ )
+
+ # Build (addr_start, addr_end, file_off, init_prot, chunk_path) ranges.
+ ranges = []
+ for cp in self._chunk_paths:
+ for m in _parse_chunk_mappings(cp):
+ if m["size"] == 0:
+ continue
+ ranges.append(
+ (
+ m["addr"],
+ m["addr"] + m["size"],
+ m["file_off"],
+ m["init_prot"],
+ cp,
+ )
+ )
+ if not ranges:
+ raise RuntimeError(
+ f"No DSC mappings parsed under {chunks_dir!r} — "
+ f"chunks present but headers unrecognised"
+ )
+
+ # Sort by addr_start for binary-search later.
+ ranges.sort(key=lambda r: r[0])
+ self._ranges = ranges
+
+ def __repr__(self):
+ return (
+ f"DSCChunks({self.chunks_dir!r}, "
+ f"{len(self._chunk_paths)} chunk(s), "
+ f"{len(self._ranges)} mapping(s), "
+ f"vm 0x{self._ranges[0][0]:X}..0x{self._ranges[-1][1]:X})"
+ )
+
+ def find_chunk_for_vma(self, vma):
+ """Return (chunk_path, file_offset) for `vma`, or None if unmapped."""
+ # Linear scan is fine; mapping count is < ~80.
+ for addr_start, addr_end, file_off, _init_prot, cp in self._ranges:
+ if addr_start <= vma < addr_end:
+ return (cp, file_off + (vma - addr_start))
+ return None
+
+ def bytes_at_vma(self, vma, length):
+ """Read `length` bytes starting at `vma`. Must be contained in one
+ mapping (we don't stitch across mappings)."""
+ loc = self.find_chunk_for_vma(vma)
+ if loc is None:
+ raise KeyError(f"vma 0x{vma:X} is not mapped by any chunk")
+ cp, foff = loc
+ with open(cp, "rb") as f:
+ f.seek(foff)
+ data = f.read(length)
+ if len(data) != length:
+ raise IOError(
+ f"short read at vma 0x{vma:X} ({len(data)} of {length})"
+ )
+ return data
+
+ def write_at_vma(self, vma, data):
+ """Persist `data` bytes at `vma` to the right chunk file. The
+ target span must lie within a single mapping."""
+ loc = self.find_chunk_for_vma(vma)
+ if loc is None:
+ raise KeyError(f"vma 0x{vma:X} is not mapped by any chunk")
+ # Verify the entire span is in this mapping.
+ end_loc = self.find_chunk_for_vma(vma + len(data) - 1)
+ if end_loc is None or end_loc[0] != loc[0]:
+ raise ValueError(
+ f"write at vma 0x{vma:X} length {len(data)} "
+ f"crosses a chunk boundary — refusing"
+ )
+ cp, foff = loc
+ with open(cp, "r+b") as f:
+ f.seek(foff)
+ f.write(data)
+
+ # ---- additional helpers used by the DSC-native canonical-site finder ----
+
+ VM_PROT_EXECUTE = 4
+
+ def mappings(self):
+ """Return [(vmaddr_start, vmaddr_end, file_offset, init_prot, chunk_path)]."""
+ return list(self._ranges)
+
+ def find_string_vmas(self, needle):
+ """Return all vmaddrs where `needle` (bytes) appears within a single
+ mapping, anchored either at the mapping start or after a NUL byte.
+
+ Restricted to executable (RX) mappings — `__TEXT` segments where
+ `__TEXT,__cstring` lives. This skips the giant LINKEDIT and
+ __DATA/__DATA_CONST mappings entirely.
+ """
+ out = []
+ for addr_start, addr_end, file_off, init_prot, cp in self._ranges:
+ if not (init_prot & self.VM_PROT_EXECUTE):
+ continue
+ size = addr_end - addr_start
+ with open(cp, "rb") as f:
+ f.seek(file_off)
+ buf = f.read(size)
+ i = 0
+ while True:
+ p = buf.find(needle, i)
+ if p < 0:
+ break
+ if p == 0 or buf[p - 1] == 0:
+ out.append(addr_start + p)
+ i = p + 1
+ return out
+
+ def iter_executable_mapping_bytes(self):
+ """Yield (chunk_path, file_offset, vmaddr_start, mapping_bytes) for
+ every mapping whose `initProt` includes VM_PROT_EXECUTE.
+
+ Non-executable mappings (LINKEDIT, __DATA, __DATA_CONST) are
+ skipped so we never try to disassemble them.
+ """
+ for addr_start, addr_end, file_off, init_prot, cp in self._ranges:
+ if not (init_prot & self.VM_PROT_EXECUTE):
+ continue
+ size = addr_end - addr_start
+ with open(cp, "rb") as f:
+ f.seek(file_off)
+ buf = f.read(size)
+ yield cp, file_off, addr_start, buf
+
+ def read_at_vma(self, vma, length, *, allow_short=False):
+ """Read `length` bytes starting at `vma`, contained in a single
+ mapping. With `allow_short=True`, returns whatever is available
+ within the containing mapping instead of raising."""
+ for addr_start, addr_end, file_off, _init_prot, cp in self._ranges:
+ if addr_start <= vma < addr_end:
+ avail = addr_end - vma
+ want = length if avail >= length else (avail if allow_short else 0)
+ if want == 0:
+ raise IOError(
+ f"read at vma 0x{vma:X} len {length} would cross "
+ f"a chunk boundary"
+ )
+ with open(cp, "rb") as f:
+ f.seek(file_off + (vma - addr_start))
+ data = f.read(want)
+ return data
+ raise KeyError(f"vma 0x{vma:X} is not mapped by any chunk")
+
+ _MH_MAGIC_64_LE = b"\xcf\xfa\xed\xfe"
+
+ def find_macho_header_before(self, vma, *, max_walk=64 * 1024 * 1024):
+ """Walk backwards from `vma` looking for a Mach-O 64 header
+ (magic 0xfeedfacf) at a page boundary. Returns the magic's
+ vmaddr, or None if not found within `max_walk` bytes inside the
+ same mapping.
+
+ Mach-O headers in the DSC are placed at page-aligned vmaddrs at
+ the start of each dylib's __TEXT segment, and an arm64e iOS
+ dylib's header is in the same mapping as its __text section, so
+ we only have to search within the containing mapping.
+
+ Implementation: read the relevant bytes from disk once and use
+ `bytes.rfind` to locate the magic, then accept only page-aligned
+ hits. Searching ~36 MiB takes <50 ms in pure Python.
+ """
+ for addr_start, addr_end, file_off, _init_prot, cp in self._ranges:
+ if not (addr_start <= vma < addr_end):
+ continue
+ local_off = vma - addr_start
+ scan_len = min(local_off, max_walk)
+ scan_start = local_off - scan_len
+ with open(cp, "rb") as f:
+ f.seek(file_off + scan_start)
+ buf = f.read(scan_len + 4) # +4 so a hit at end is captured
+ i = len(buf)
+ while True:
+ found = buf.rfind(self._MH_MAGIC_64_LE, 0, i)
+ if found < 0:
+ return None
+ candidate_vma = addr_start + scan_start + found
+ if (candidate_vma & 0xFFF) == 0:
+ return candidate_vma
+ # Keep searching backwards.
+ i = found
+ # unreachable
+ return None
+
+ def read_install_name_at(self, header_vma):
+ """Given the vmaddr of a Mach-O 64 header in the DSC, return the
+ dylib's install name (LC_ID_DYLIB), or None if the file isn't a
+ dylib or the LC isn't present.
+ """
+ # Read the first 64KB of LCs (more than enough).
+ try:
+ head = self.read_at_vma(header_vma, 64 * 1024, allow_short=True)
+ except (KeyError, IOError):
+ return None
+ if struct.unpack_from("<I", head, 0)[0] != 0xFEEDFACF:
+ return None
+ ncmds = struct.unpack_from("<I", head, 16)[0]
+ sizeofcmds = struct.unpack_from("<I", head, 20)[0]
+ # Sanity-bound the LC walk.
+ if ncmds > 4096 or sizeofcmds > 64 * 1024:
+ return None
+ o = 32
+ for _ in range(ncmds):
+ if o + 8 > len(head):
+ return None
+ cmd, cmdsize = struct.unpack_from("<II", head, o)
+ if cmdsize < 8 or cmdsize > 64 * 1024:
+ return None
+ if cmd == 0xD: # LC_ID_DYLIB
+ # struct dylib_command { cmd, cmdsize, dylib { name(uint32 offset),
+ # timestamp, current_version, compatibility_version } }
+ name_off_in_cmd = struct.unpack_from("<I", head, o + 8)[0]
+ name_pos = o + name_off_in_cmd
+ if name_pos >= len(head) or name_pos >= o + cmdsize:
+ return None
+ end = head.find(b"\x00", name_pos, o + cmdsize)
+ if end < 0:
+ return None
+ try:
+ return head[name_pos:end].decode("utf-8")
+ except UnicodeDecodeError:
+ return head[name_pos:end].decode("latin-1")
+ o += cmdsize
+ if o > sizeofcmds + 32:
+ return None
+ return None
diff --git a/scripts/patchers/cfw_dsc_codesign.py b/scripts/patchers/cfw_dsc_codesign.py
new file mode 100644
index 00000000..4390781b
--- /dev/null
+++ b/scripts/patchers/cfw_dsc_codesign.py
@@ -0,0 +1,294 @@
+"""DSC code-signature page-hash re-attestation.
+
+Background
+----------
+On iPhone17,3 / iOS 26.1 with TXM (`codeSigningMonitor == 2`) enabled,
+the kernel defers per-page hash validation to TXM (see source comments
+in `research/reference/xnu/osfmk/vm/vm_fault.c:2763-2780`). TXM has
+the original slot hashes baked in via the registered DSC signature, so
+any byte-mangle inside an executable mapping causes
+`KERN_PROTECTION_FAILURE` / `CODESIGNING / Invalid Page` SIGKILL on
+the first demand-page-in of the affected 16 KiB page. (An earlier
+XNU-side `_cs_validate_page` short-circuit kernel patch was attempted
+and shipped, then removed — it never fires on this hardware because
+the CSM-defer gate in `vm_fault_cs_check_violation` runs first.)
+
+To make a byte-mangled DSC dylib executable again, we recompute the
+SHA-256 slot hash for the 16 KiB page that contains the modification
+and overwrite the matching entry in the chunk's `CS_CodeDirectory`.
+
+DSC code-signature layout (verified empirically on
+`iPhone17,3_26.1_23B85` chunked DSC):
+
+ Each chunk file has a `dyld_cache_header`:
+ u64 codeSignatureOffset @ 0x28 (chunk-file relative)
+ u64 codeSignatureSize @ 0x30
+
+ At `codeSignatureOffset` there's a `CS_SuperBlob` (big-endian):
+ u32 magic = 0xFADE0CC0
+ u32 length
+ u32 count
+ CS_BlobIndex blobs[count]:
+ u32 type
+ u32 offset (within SuperBlob)
+
+ The CSSLOT_CODEDIRECTORY entry (type==0) points at a
+ `CS_CodeDirectory` blob (also big-endian):
+ u32 magic = 0xFADE0C02
+ u32 length
+ u32 version
+ u32 flags
+ u32 hashOffset (within blob — to slot[0])
+ u32 identOffset
+ u32 nSpecialSlots
+ u32 nCodeSlots
+ u32 codeLimit (chunk-file byte range covered)
+ u8 hashSize (32 for SHA-256)
+ u8 hashType (2 for SHA-256)
+ u8 platform
+ u8 pageSize (log2 — 14 == 16 KiB)
+ u32 spare2
+ ... (more in later versions)
+
+ Slot N stores SHA-256 of chunk bytes [N*pageSize, (N+1)*pageSize)
+ and lives at chunk-file offset:
+ codeSignatureOffset + cd_offset_in_superblob
+ + hashOffset + N * hashSize
+
+DSC dylibs do NOT carry their own LC_CODE_SIGNATURE — only the
+chunk-level signature applies. Page indexing is against the chunk
+file offset, not the dylib's __TEXT offset.
+
+CDHash side effect (empirically benign)
+---------------------------------------
+Modifying slot hashes changes the CodeDirectory's contents, which
+changes its CDHash (SHA-256 of the CD blob). The existing JB kernel
+patch `patch_amfi_cdhash_in_trustcache` bypasses AMFI's per-image
+CDHash trust-cache lookup, and TXM accepts the modified-CDHash DSC
+at cache-load time on iPhone17,3 / iOS 26.1 (verified: device boots
+with re-attested chunks). If a future build introduces a stricter
+CDHash check at DSC mount, the next step would be a TXM-side bypass
+at cache load — at the time this module was written that wasn't
+needed.
+
+"""
+
+import hashlib
+import os
+import struct
+
+
+# CS constants.
+CSMAGIC_EMBEDDED_SIGNATURE = 0xFADE0CC0
+CSMAGIC_CODEDIRECTORY = 0xFADE0C02
+CSSLOT_CODEDIRECTORY = 0
+
+CS_HASHTYPE_SHA256 = 2
+
+
+def _read_chunk_cd_blob(chunk_path):
+ """Locate the chunk's CS_CodeDirectory.
+
+ Returns a dict with the fields we need to write back slot hashes:
+ cd_file_off (chunk-file offset of the CD blob)
+ cd_length (total length of the CD blob in bytes)
+ hash_offset (offset within the CD blob to slot[0])
+ hash_size (bytes per slot — 32 for SHA-256)
+ n_code_slots (number of slots)
+ code_limit (chunk byte range covered)
+ page_size (1 << pageSize_log2 — typically 16384)
+
+ Returns None if the chunk file isn't recognised (header magic
+ mismatch, no CodeDirectory in SuperBlob, etc.).
+ """
+ with open(chunk_path, "rb") as f:
+ head = f.read(0x100)
+ if not head.startswith(b"dyld_"):
+ return None
+
+ # codeSignatureOffset at 0x28, codeSignatureSize at 0x30 (both u64 LE).
+ cs_off, cs_size = struct.unpack_from("<QQ", head, 0x28)
+ if cs_off == 0 or cs_size == 0:
+ return None
+
+ # SuperBlob (big-endian).
+ with open(chunk_path, "rb") as f:
+ f.seek(cs_off)
+ sb_head = f.read(12)
+ if len(sb_head) != 12:
+ return None
+ sb_magic, sb_length, sb_count = struct.unpack(">III", sb_head)
+ if sb_magic != CSMAGIC_EMBEDDED_SIGNATURE:
+ return None
+ if sb_length > cs_size or sb_count > 256:
+ return None
+
+ # Read the BlobIndex array.
+ with open(chunk_path, "rb") as f:
+ f.seek(cs_off + 12)
+ idx_bytes = f.read(sb_count * 8)
+ if len(idx_bytes) != sb_count * 8:
+ return None
+
+ cd_off_in_sb = None
+ for i in range(sb_count):
+ slot_type, slot_off = struct.unpack_from(">II", idx_bytes, i * 8)
+ if slot_type == CSSLOT_CODEDIRECTORY:
+ cd_off_in_sb = slot_off
+ break
+ if cd_off_in_sb is None:
+ return None
+
+ cd_file_off = cs_off + cd_off_in_sb
+
+ # CD blob header (big-endian).
+ with open(chunk_path, "rb") as f:
+ f.seek(cd_file_off)
+ cd_head = f.read(44)
+ if len(cd_head) != 44:
+ return None
+ cd_magic, cd_length, _version, _flags = struct.unpack_from(">IIII", cd_head, 0)
+ if cd_magic != CSMAGIC_CODEDIRECTORY:
+ return None
+ hash_offset, _ident_offset, _n_special, n_code_slots, code_limit = \
+ struct.unpack_from(">IIIII", cd_head, 16)
+ hash_size = cd_head[36]
+ hash_type = cd_head[37]
+ page_size_log2 = cd_head[39]
+ if hash_type != CS_HASHTYPE_SHA256 or hash_size != 32:
+ # We only know how to recompute SHA-256 slots. Other hash types
+ # (SHA-1, SHA-256-truncated) would need additional code paths.
+ return None
+ page_size = 1 << page_size_log2
+ if page_size == 0 or page_size > (1 << 20):
+ return None
+
+ return {
+ "cd_file_off": cd_file_off,
+ "cd_length": cd_length,
+ "hash_offset": hash_offset,
+ "hash_size": hash_size,
+ "n_code_slots": n_code_slots,
+ "code_limit": code_limit,
+ "page_size": page_size,
+ }
+
+
+def reattest_modified_pages(chunks, modified_vmas, *, dry_run=False, verbose=True):
+ """Recompute and write SHA-256 slot hashes for every 16 KiB page
+ that contains a `modified_vmas` entry.
+
+ `chunks` a DSCChunks instance.
+ `modified_vmas` iterable of vmaddrs we wrote to. Each is reduced
+ to its containing chunk + page; duplicate pages
+ are coalesced.
+ `dry_run` if True, log only, do not write.
+
+ Returns a list of dicts (one per page recomputed) for diagnostics:
+ {chunk_path, page_index, chunk_off, sha256_before, sha256_after}
+ """
+ # Group vmas by chunk → set of page indices to recompute.
+ chunks_pages = {} # chunk_path → set(page_index)
+ chunk_meta = {} # chunk_path → cd metadata (lazy-loaded)
+
+ for vma in modified_vmas:
+ loc = chunks.find_chunk_for_vma(vma)
+ if loc is None:
+ if verbose:
+ print(f" [-] re-attest: vma 0x{vma:X} not mapped in any chunk")
+ continue
+ cp, foff = loc
+ meta = chunk_meta.get(cp)
+ if meta is None:
+ meta = _read_chunk_cd_blob(cp)
+ if meta is None:
+ if verbose:
+ print(f" [-] re-attest: chunk {os.path.basename(cp)!r} "
+ f"has no recognised CS_SuperBlob/CD — skipping")
+ # Cache the None so we don't re-probe.
+ chunk_meta[cp] = False
+ continue
+ chunk_meta[cp] = meta
+ elif meta is False:
+ continue
+ page_size = meta["page_size"]
+ page_index = foff // page_size
+ if page_index >= meta["n_code_slots"]:
+ if verbose:
+ print(f" [-] re-attest: vma 0x{vma:X} -> page {page_index} "
+ f"is past nCodeSlots ({meta['n_code_slots']}) in "
+ f"{os.path.basename(cp)} — skipping")
+ continue
+ if (page_index + 1) * page_size > meta["code_limit"]:
+ if verbose:
+ print(f" [-] re-attest: vma 0x{vma:X} -> page {page_index} "
+ f"overruns codeLimit (0x{meta['code_limit']:X}) in "
+ f"{os.path.basename(cp)} — skipping")
+ continue
+ chunks_pages.setdefault(cp, set()).add(page_index)
+
+ if not chunks_pages:
+ if verbose:
+ print(f" [.] re-attest: no eligible pages")
+ return []
+
+ diagnostics = []
+ total = 0
+ for cp, pages in chunks_pages.items():
+ meta = chunk_meta[cp]
+ cd_file_off = meta["cd_file_off"]
+ hash_offset = meta["hash_offset"]
+ hash_size = meta["hash_size"]
+ page_size = meta["page_size"]
+
+ # Open once per chunk for read/write.
+ mode = "rb" if dry_run else "r+b"
+ with open(cp, mode) as f:
+ for page_index in sorted(pages):
+ page_off = page_index * page_size
+ slot_off = cd_file_off + hash_offset + page_index * hash_size
+
+ f.seek(page_off)
+ page_data = f.read(page_size)
+ if len(page_data) != page_size:
+ if verbose:
+ print(f" [-] re-attest: short read at "
+ f"page {page_index} of {os.path.basename(cp)} "
+ f"(got {len(page_data)} of {page_size})")
+ continue
+ new_hash = hashlib.sha256(page_data).digest()
+
+ f.seek(slot_off)
+ old_hash = f.read(hash_size)
+
+ if old_hash == new_hash:
+ if verbose:
+ print(f" [.] re-attest: page {page_index} of "
+ f"{os.path.basename(cp)} slot already matches "
+ f"(no-op)")
+ continue
+
+ if not dry_run:
+ f.seek(slot_off)
+ f.write(new_hash)
+
+ if verbose:
+ action = "would write" if dry_run else "wrote"
+ print(f" [+] re-attest: {action} slot {page_index} of "
+ f"{os.path.basename(cp)} "
+ f"({old_hash.hex()[:8]}.. -> {new_hash.hex()[:8]}..)")
+ diagnostics.append({
+ "chunk_path": cp,
+ "page_index": page_index,
+ "chunk_off": page_off,
+ "slot_off": slot_off,
+ "sha256_before": old_hash.hex(),
+ "sha256_after": new_hash.hex(),
+ })
+ total += 1
+
+ if verbose:
+ action = "would update" if dry_run else "updated"
+ print(f" [+] re-attest: {action} {total} slot hash(es) across "
+ f"{len(chunks_pages)} chunk(s)")
+ return diagnostics
diff --git a/scripts/patchers/cfw_patch_hv_vmm.py b/scripts/patchers/cfw_patch_hv_vmm.py
new file mode 100644
index 00000000..9a4e881e
--- /dev/null
+++ b/scripts/patchers/cfw_patch_hv_vmm.py
@@ -0,0 +1,142 @@
+"""hv_vmm_present user-mode patch module — cstring-mangle approach.
+
+Mangles the cstring "kern.hv_vmm_present\\0" in targeted user-mode
+binaries so they query a kernel sysctl with a different name. On a
+kernel patched by `KernelJBPatchHvVmmRename` (which renames the OID
+from `hv_vmm_present` to `Xv_vmm_present`), the new design is:
+
+ - Caller queries `kern.hv_vmm_present` (UNPATCHED binary):
+ kernel returns ENOENT. The canonical defensive post-call check
+ (`cbnz w0, skip` / `cmp w0,#0 ; b.ne skip`) takes the skip
+ path; the cached `is_vmm` byte stays at BSS-zero (0). The
+ binary thinks it is NOT running in a VM.
+
+ - Caller queries `kern.Xv_vmm_present` (PATCHED binary):
+ kernel returns 1 (the real value of the OID's int variable).
+ The defensive check passes; the cache stores 1. The binary
+ thinks it IS running in a VM — same as a stock device.
+
+Patch shape (per cstring occurrence): byte 5 of the cstring
+`'h'` (0x68) → `'X'` (0x58). The mangled cstring is
+`"kern.Xv_vmm_present\\0"`. Byte 5 is chosen so the `kern.` top-level
+sysctl namespace is preserved — without that, the kernel's
+name-to-MIB resolver wouldn't route the call to any OID.
+
+Public entrypoints:
+ patch_hv_vmm(filepath, *, dry_run=False) -> int
+ Patch a standalone Mach-O. Returns the number of cstring sites
+ mangled.
+
+ find_string_sites(data: bytes) -> list[dict]
+ Find every cstring occurrence in an in-memory Mach-O.
+"""
+
+import struct
+
+from .cfw_asm import parse_macho_sections, _log_asm
+
+
+NEEDLE = b"kern.hv_vmm_present\x00"
+# Mangle byte offset (0-based within NEEDLE). Position 5 is the 'h' of
+# "hv_vmm_present" — the first byte after the "kern." namespace prefix.
+# Keeping the "kern." prefix intact ensures the kernel's sysctl name-to-MIB
+# resolver routes the call to a real OID (registered as `Xv_vmm_present`
+# under `kern` by the companion kernel patch). Byte 0 would have produced
+# `Xern.hv_vmm_present`, which can never resolve because `Xern` isn't a
+# registered top-level sysctl namespace.
+MANGLE_OFFSET = 5
+ORIGINAL_BYTE = b"h" # 0x68
+MANGLED_BYTE = b"X" # 0x58
+MANGLED_NEEDLE = NEEDLE[:MANGLE_OFFSET] + MANGLED_BYTE + NEEDLE[MANGLE_OFFSET + 1:]
+
+
+def find_string_sites(data):
+ """Return all unmangled "kern.hv_vmm_present\\0" cstring occurrences.
+
+ Each entry: {string_vma, file_offset, section}.
+
+ Hits are anchored at a cstring boundary (preceded by a NUL byte or
+ at the start of the section) so partial-match substrings inside
+ longer cstrings won't be returned.
+ """
+ sections = parse_macho_sections(data)
+ out = []
+ for sec_name, (vma, size, foff) in sections.items():
+ _, _, sect = sec_name.partition(",")
+ # __cstring is where the linker puts unique C-string literals.
+ # Some method-name / class-name pools could in principle hold
+ # the same bytes too; we include them for safety.
+ if sect not in ("__cstring", "__objc_methname", "__objc_classname"):
+ continue
+ buf = bytes(data[foff:foff + size])
+ i = 0
+ while True:
+ p = buf.find(NEEDLE, i)
+ if p < 0:
+ break
+ if p == 0 or buf[p - 1] == 0:
+ out.append(
+ {
+ "string_vma": vma + p,
+ "file_offset": foff + p,
+ "section": sec_name,
+ }
+ )
+ i = p + 1
+ return out
+
+
+def is_already_mangled(data):
+ """Detect whether the binary already contains the mangled form.
+
+ Used purely for friendlier logging — `find_string_sites` already
+ returns empty on a mangled binary, so the patch flow is idempotent
+ regardless.
+ """
+ return bytes(data).find(MANGLED_NEEDLE) >= 0
+
+
+def patch_hv_vmm(filepath, *, dry_run=False):
+ """Mangle the kern.hv_vmm_present cstring in a standalone Mach-O.
+
+ Returns the count of mangled cstring sites. Idempotent.
+ """
+ data = bytearray(open(filepath, "rb").read())
+ sites = find_string_sites(bytes(data))
+ if not sites:
+ if is_already_mangled(bytes(data)):
+ print(f" [.] {filepath}: already mangled (no original "
+ f"'kern.hv_vmm_present' cstring present)")
+ else:
+ print(f" [.] {filepath}: 'kern.hv_vmm_present' cstring not "
+ f"present — nothing to do")
+ return 0
+
+ print(f" [+] {len(sites)} cstring occurrence(s) in {filepath}")
+ n = 0
+ for s in sites:
+ foff = s["file_offset"]
+ original = bytes(data[foff:foff + len(NEEDLE)])
+ # Idempotence: if byte at MANGLE_OFFSET is already mangled, skip.
+ if original[MANGLE_OFFSET:MANGLE_OFFSET + 1] == MANGLED_BYTE:
+ print(f" [.] string@0x{s['string_vma']:X} already mangled "
+ f"(byte {MANGLE_OFFSET} is already {MANGLED_BYTE.decode()!r})")
+ continue
+ if original != NEEDLE:
+ print(f" [-] string@0x{s['string_vma']:X} bytes look unexpected "
+ f"({original!r}); skipping")
+ continue
+ print(f" patching string@0x{s['string_vma']:X} (sect={s['section']}): "
+ f"byte {MANGLE_OFFSET} {ORIGINAL_BYTE.decode()!r} -> "
+ f"{MANGLED_BYTE.decode()!r} "
+ f"('kern.hv_vmm_present' -> 'kern.Xv_vmm_present')")
+ data[foff + MANGLE_OFFSET:foff + MANGLE_OFFSET + 1] = MANGLED_BYTE
+ n += 1
+
+ if dry_run:
+ print(f" [.] dry-run — not writing back")
+ return n
+ if n > 0:
+ open(filepath, "wb").write(data)
+ print(f" [+] {filepath}: mangled {n} cstring occurrence(s)")
+ return n
diff --git a/scripts/patchers/cfw_patch_hv_vmm_dsc.py b/scripts/patchers/cfw_patch_hv_vmm_dsc.py
new file mode 100644
index 00000000..415f177c
--- /dev/null
+++ b/scripts/patchers/cfw_patch_hv_vmm_dsc.py
@@ -0,0 +1,266 @@
+"""DSC-side orchestrator for the hv_vmm_present user-mode patch
+(blacklist-flip design).
+
+Design summary
+--------------
+This module works in tandem with `KernelJBPatchHvVmmRename` (Swift
+kernel patcher) which renames the kernel's sysctl OID from
+`hv_vmm_present` to `Xv_vmm_present`. After the kernel patch:
+
+ - `sysctlbyname("kern.hv_vmm_present", ...)` returns ENOENT.
+ - `sysctlbyname("kern.Xv_vmm_present", ...)` returns 1 (the OID's
+ real int value, since this device is in fact a VM).
+
+This module's job is to selectively rewrite cstrings inside DSC
+dylibs so each dylib queries either the original (now ENOENT-returning)
+name or the new (truthful-1-returning) name. Byte 5 of the
+`"kern.hv_vmm_present\\0"` cstring is flipped from `'h'` (0x68) to
+`'X'` (0x58), producing `"kern.Xv_vmm_present\\0"` — the same name the
+kernel OID now answers to. The `kern.` namespace prefix is preserved
+(byte-0 mangle would have produced `Xern.hv_vmm_present` which can
+never resolve because `Xern` isn't a registered top-level).
+
+Blacklist semantics
+-------------------
+The `DONT_PATCH_INSTALL_NAMES` list names dylibs whose cstring stays
+ORIGINAL — they query `kern.hv_vmm_present`, get ENOENT, and the
+caller's defensive post-call check leaves the cached `is_vmm` byte
+at BSS-zero (0). Those libs end up thinking "not in a VM" — which is
+what we want for sign-in / device-likeness consumers.
+
+Every OTHER dylib that contains the cstring gets the byte-5 mangle —
+they query `kern.Xv_vmm_present`, get 1, cache 1, and think "in a VM"
+— exactly the value they would have seen on a stock device. This
+keeps the graphics path (libMobileGestalt, PhotoFoundation,
+AirPlaySupport, VisionKitCore, CoreVideo) and compute/accel fast
+paths (CoreML, Espresso, AppleNeuralEngine, …) intact, because they
+keep getting their original answer.
+
+"""
+
+import os
+import struct
+import sys
+
+from .cfw_patch_hv_vmm import (
+ NEEDLE,
+ MANGLED_NEEDLE,
+ MANGLE_OFFSET,
+ ORIGINAL_BYTE,
+ MANGLED_BYTE,
+)
+from .cfw_dsc_chunks import DSCChunks
+from .cfw_dsc_codesign import reattest_modified_pages
+
+
+# ─────────────────────────────────────────────────────────────────────
+# Blacklist of DSC dylibs to LEAVE UNPATCHED.
+#
+# A dylib in this list keeps its cstring as `"kern.hv_vmm_present\0"`.
+# With the kernel-rename patch in place, that name resolves to ENOENT,
+# the dylib's defensive `cbnz w0, skip` takes the skip path, and the
+# cached `is_vmm` byte stays at BSS-zero (0). The dylib thinks the
+# device is not a VM.
+#
+# A dylib NOT in this list has its cstring rewritten to
+# `"kern.Xv_vmm_present\0"` (byte 5 mangle, 'h' → 'X'). That name
+# resolves to the renamed OID on the kernel side and returns 1. The
+# dylib thinks the device IS a VM — same as stock.
+#
+# Comment a single line out (re-enabling the patch for that dylib —
+# i.e., MOVING IT OUT of the blacklist) to bisect which consumer
+# regressed an observable.
+#
+# Order is grouped by why each entry needs to lie about VM presence;
+# the order itself doesn't affect patching.
+# ─────────────────────────────────────────────────────────────────────
+DONT_PATCH_INSTALL_NAMES = (
+ # ── Identity / activation / anti-abuse.
+ "/System/Library/PrivateFrameworks/AAAFoundation.framework/AAAFoundation",
+ "/System/Library/PrivateFrameworks/AuthKit.framework/AuthKit",
+ "/System/Library/PrivateFrameworks/IDSFoundation.framework/IDSFoundation",
+ "/System/Library/PrivateFrameworks/DeviceIdentity.framework/DeviceIdentity",
+ "/System/Library/PrivateFrameworks/DeviceCheckInternal.framework/DeviceCheckInternal",
+ "/System/Library/PrivateFrameworks/MobileActivation.framework/MobileActivation",
+ "/System/Library/PrivateFrameworks/ApplePushService.framework/ApplePushService",
+
+ # ── Store / IAP.
+ "/System/Library/PrivateFrameworks/AppStoreUtilities.framework/AppStoreUtilities",
+
+ # ── Consumer services.
+ "/System/Library/PrivateFrameworks/CorePrescription.framework/CorePrescription",
+ "/System/Library/PrivateFrameworks/CoreCDP.framework/CoreCDP",
+ "/System/Library/PrivateFrameworks/EmailFoundation.framework/EmailFoundation",
+ "/System/Library/PrivateFrameworks/FindMyBase.framework/FindMyBase",
+ "/System/Library/PrivateFrameworks/TrialServer.framework/TrialServer",
+ "/System/Library/PrivateFrameworks/DVTInstrumentsUtilities.framework/DVTInstrumentsUtilities",
+ "/System/Library/PrivateFrameworks/WatchdogServiceManagement.framework/WatchdogServiceManagement",
+)
+_DONT_PATCH_SET: frozenset[str] = frozenset(DONT_PATCH_INSTALL_NAMES)
+
+
+def _classify(chunks, string_vma):
+ """Walk back from `string_vma` to find the containing Mach-O header
+ and return its install name (LC_ID_DYLIB), or None if not found.
+ """
+ header_vma = chunks.find_macho_header_before(string_vma)
+ if header_vma is None:
+ return None
+ return chunks.read_install_name_at(header_vma)
+
+
+def patch_hv_vmm_in_dsc(chunks_dir, *, dry_run=False):
+ """Apply the blacklist-flip mangle inside the DSC chunks at
+ `chunks_dir`.
+
+ For every occurrence of `"kern.hv_vmm_present\\0"` in any
+ executable DSC mapping, resolve the containing dylib's install
+ name; if the install name is NOT in `DONT_PATCH_INSTALL_NAMES`,
+ rewrite byte 5 of the cstring from 'h' to 'X'.
+
+ Returns {install_name: count_of_cstring_sites_mangled}.
+ """
+ chunks = DSCChunks(chunks_dir)
+ print(f" [.] {chunks!r}")
+
+ print(f" [.] locating cstring \"kern.hv_vmm_present\\0\"...")
+ string_vmas = chunks.find_string_vmas(NEEDLE)
+ # Also pick up cstrings already mangled by a prior run — we still
+ # need their 16 KiB pages added to the re-attestation set so a
+ # re-run on a previously-patched DSC keeps slot hashes in sync.
+ already_mangled_vmas = chunks.find_string_vmas(MANGLED_NEEDLE)
+ if not string_vmas and not already_mangled_vmas:
+ print(f" [-] cstring not present in any executable mapping; "
+ f"nothing to do (either patched already or absent)")
+ return {}
+ print(f" [.] {len(string_vmas)} pristine + {len(already_mangled_vmas)} "
+ f"already-mangled cstring occurrence(s) found")
+
+ print(f" [.] resolving containing dylib for each occurrence "
+ f"(blacklist has {len(_DONT_PATCH_SET)} entries — these stay "
+ f"unpatched)...")
+ results = {}
+ patched_total = 0
+ skipped_in_blacklist = 0
+ skipped_unclassified = 0
+ refused = 0
+ # Track every vmaddr we actually mangled so the re-attestation pass
+ # at the end recomputes exactly the affected 16 KiB pages.
+ modified_vmas: list[int] = []
+
+ for vma in sorted(string_vmas):
+ try:
+ install_name = _classify(chunks, vma)
+ except Exception as e:
+ print(f" [-] could not classify string@0x{vma:X}: "
+ f"{type(e).__name__}: {e}")
+ install_name = None
+ label = install_name or "<unknown dylib>"
+
+ if install_name is None:
+ # Unknown dylibs are NOT in the blacklist → would normally
+ # be patched. Refuse here because we can't audit them.
+ print(f" [.] SKIP (no install name resolvable) "
+ f"string@0x{vma:X}")
+ skipped_unclassified += 1
+ results[label] = results.get(label, 0)
+ continue
+
+ if install_name in _DONT_PATCH_SET:
+ print(f" [.] SKIP (blacklisted — stays unpatched, "
+ f"will hit ENOENT): {install_name} string@0x{vma:X}")
+ skipped_in_blacklist += 1
+ results[label] = results.get(label, 0)
+ continue
+
+ # Sanity: the byte at vma must currently look like the pristine cstring.
+ try:
+ current = chunks.read_at_vma(vma, len(NEEDLE))
+ except (KeyError, IOError) as e:
+ print(f" [-] read failed at 0x{vma:X}: {e}")
+ refused += 1
+ continue
+ if current[MANGLE_OFFSET:MANGLE_OFFSET + 1] == MANGLED_BYTE:
+ # Already mangled by a prior run; nothing to write, but
+ # the re-attest pass below picks it up via already_mangled_vmas.
+ print(f" [.] already mangled at byte {MANGLE_OFFSET}: "
+ f"{label} string@0x{vma:X}")
+ results[label] = results.get(label, 0)
+ continue
+ if current != NEEDLE:
+ print(f" [-] unexpected bytes at 0x{vma:X} "
+ f"({current!r}); refusing ({label})")
+ refused += 1
+ continue
+
+ action = "would mangle" if dry_run else "mangled"
+ if not dry_run:
+ # Write the single byte at vma + MANGLE_OFFSET.
+ chunks.write_at_vma(vma + MANGLE_OFFSET, MANGLED_BYTE)
+ print(f" [+] {action} {label} string@0x{vma:X} "
+ f"byte {MANGLE_OFFSET} "
+ f"({ORIGINAL_BYTE.decode()} -> {MANGLED_BYTE.decode()}) "
+ f"now queries kern.Xv_vmm_present, will see 1 (in a VM)")
+ results[label] = results.get(label, 0) + 1
+ patched_total += 1
+ modified_vmas.append(vma)
+
+ # Second pass: previously-mangled cstrings. Add their pages to the
+ # re-attestation set so a re-run brings slot hashes into sync.
+ # IMPORTANT: we INTENTIONALLY do not refuse a mangled cstring whose
+ # install_name is in the blacklist — that's an out-of-band action
+ # (someone re-enabled patching for that dylib in a prior run and
+ # then commented it back into the blacklist). We log it loudly.
+ reattest_only_count = 0
+ blacklist_drift_count = 0
+ for vma in sorted(already_mangled_vmas):
+ try:
+ install_name = _classify(chunks, vma)
+ except Exception:
+ install_name = None
+ if install_name is None:
+ continue
+ if install_name in _DONT_PATCH_SET:
+ # This dylib is supposed to be unpatched but appears mangled.
+ # Most likely the operator mangled and then re-added it to
+ # the blacklist; we surface this and still re-attest the
+ # page so the slot hash matches the current bytes.
+ print(f" [!] drift: {install_name} is in the blacklist "
+ f"but already mangled at string@0x{vma:X} — slot will "
+ f"be re-attested to current bytes; consider whether "
+ f"you actually want this dylib re-patched")
+ modified_vmas.append(vma)
+ blacklist_drift_count += 1
+ continue
+ modified_vmas.append(vma)
+ reattest_only_count += 1
+ if reattest_only_count:
+ print(f" [.] also queueing {reattest_only_count} already-mangled "
+ f"cstring page(s) for re-attestation")
+ if blacklist_drift_count:
+ print(f" [.] {blacklist_drift_count} blacklisted dylib(s) found "
+ f"mangled on disk — see drift warnings above")
+
+ # Page-hash re-attestation. On hardware with TXM enforcement
+ # (`codeSigningMonitor == 2` — iPhone17,3 / iOS 26.1), the kernel
+ # defers per-page validation to TXM (vm_fault.c:2763-2780). TXM has
+ # the original SHA-256 slot hashes baked in via the DSC's chunk-
+ # level CS_CodeDirectory, so a byte-mangle alone causes
+ # `KERN_PROTECTION_FAILURE` / `CODESIGNING / Invalid Page` SIGKILL
+ # of any process that demand-pages the patched page. Recomputing
+ # the slot hash makes the modified (content, hash) pair consistent.
+ # The CDHash of the CD blob changes as a side-effect but TXM
+ # accepts it at DSC mount time on this build (verified empirically).
+ if modified_vmas and not dry_run:
+ print(f" [.] re-attesting {len(modified_vmas)} modified page(s)...")
+ reattest_modified_pages(chunks, modified_vmas, dry_run=False)
+ elif modified_vmas and dry_run:
+ print(f" [.] dry-run: would re-attest {len(modified_vmas)} "
+ f"modified page(s)")
+ reattest_modified_pages(chunks, modified_vmas, dry_run=True)
+
+ print(f" [+] DSC patch complete: {patched_total} cstring(s) mangled, "
+ f"{skipped_in_blacklist} in-blacklist (left unpatched), "
+ f"{skipped_unclassified} unclassified, "
+ f"{refused} refused/error")
+ return results
diff --git a/scripts/patchers/cfw_patch_hv_vmm_rootfs.py b/scripts/patchers/cfw_patch_hv_vmm_rootfs.py
new file mode 100644
index 00000000..d23c6cb6
--- /dev/null
+++ b/scripts/patchers/cfw_patch_hv_vmm_rootfs.py
@@ -0,0 +1,216 @@
+"""Rootfs-side orchestrator for the hv_vmm_present user-mode patch
+(blacklist-flip design, standalone-binary side).
+
+Companion to:
+ - The kernel-side `KernelJBPatchHvVmmRename` (renames the OID name
+ cstring from `hv_vmm_present` to `Xv_vmm_present` AND mangles
+ every kernel-internal `\\0kern.hv_vmm_present\\0` cstring inside
+ kexts so they query the new name and keep seeing 1).
+ - The DSC-side `cfw_patch_hv_vmm_dsc.py` (same blacklist-flip,
+ mangles every DSC dylib's cstring EXCEPT those in
+ `DONT_PATCH_INSTALL_NAMES`).
+
+This module covers the third domain: standalone Mach-O binaries on
+the device's rootfs that hard-code `"kern.hv_vmm_present\\0"` in their
+own `__TEXT,__cstring`. Under the new design every such binary needs
+to either:
+
+ - Be MANGLED (byte 5 of the cstring: 'h' -> 'X') so it queries
+ `kern.Xv_vmm_present`, gets the truthful `1` back from the
+ renamed OID, and behaves as it would on a real device. Default
+ behaviour: most platform daemons (watchdogd, mobile_obliterator,
+ DumpPanic, …) need this — without it they hit the renamed OID's
+ ENOENT, interpret as "not in a VM", and take real-hardware code
+ paths that don't exist on the VM (concrete repro: watchdogd
+ SIGTRAP → kernel panic with `watchdogd[N] exited`).
+
+ - Be LEFT ALONE (`DONT_PATCH_ROOTFS_PATHS` blacklist) so its
+ cstring still says `kern.hv_vmm_present`, that query returns
+ ENOENT after the kernel rename, the binary's defensive
+ post-call check leaves the cached `is_vmm` byte at BSS-zero (0),
+ and the binary thinks "not in a VM". Use this for the sign-in
+ / activation / Apple-ID stack — the same role intent we apply
+ to the DSC frameworks they link.
+
+Data sources for ALL_KNOWN_ROOTFS_PATHS
+---------------------------------------
+The list below is taken verbatim from a `grep -rla 'kern\\.hv_vmm_present'
+ipsws/iPhone17,3_26.1_23B85_Restore_extracted/ --exclude-dir='Cryptexes'`
+on the extracted rootfs of iPhone17,3 26.1 23B85, minus
+`/System/Library/Caches/com.apple.kernelcaches/kernelcache.release.iphone17`
+which is the kernel itself and is patched by the firmware patcher
+not this module. If future iOS builds add new daemons that reference
+the cstring, re-run that grep and add the new paths here.
+"""
+
+import sys
+
+
+# ─────────────────────────────────────────────────────────────────────
+# Complete list of rootfs Mach-Os that contain the cstring
+# `kern.hv_vmm_present\0` somewhere in their __TEXT,__cstring.
+#
+# Established once by scanning the extracted rootfs of iPhone17,3 /
+# iOS 26.1 / 23B85. Re-derive when porting to a new iOS build by:
+# grep -rla 'kern\.hv_vmm_present' \
+# ipsws/<new>_Restore_extracted/ \
+# --exclude-dir='Cryptexes' 2>/dev/null
+# ─────────────────────────────────────────────────────────────────────
+ALL_KNOWN_ROOTFS_PATHS: tuple[str, ...] = (
+ # /usr/libexec daemons
+ "/usr/libexec/adid",
+ "/usr/libexec/DumpPanic",
+ "/usr/libexec/fairplaydeviceidentityd",
+ "/usr/libexec/mmaintenanced",
+ "/usr/libexec/modelmanagerd",
+ "/usr/libexec/mobile_obliterator",
+ "/usr/libexec/mobileactivationd",
+ "/usr/libexec/networkserviceproxy",
+ "/usr/libexec/promotedcontentd",
+ "/usr/libexec/storagekitd",
+ "/usr/libexec/terminusd",
+ "/usr/libexec/transparencyd",
+ "/usr/libexec/trustd",
+ "/usr/libexec/watchdogd",
+
+ # /usr/sbin
+ "/usr/sbin/absd",
+ "/usr/sbin/bluetoothd",
+ "/usr/sbin/fairplayd.H2",
+
+ # /usr/lib
+ "/usr/lib/libMobileGestalt.dylib",
+
+ # /System/Library/CoreServices
+ "/System/Library/CoreServices/ClarityBoard.app/ClarityBoard",
+
+ # /System/Library/DataClassMigrators
+ "/System/Library/DataClassMigrators/MobileActivationMigrator.migrator/MobileActivationMigrator",
+
+ # /System/Library/ExtensionKit
+ "/System/Library/ExtensionKit/Extensions/HostInferenceProviderService.appex/HostInferenceProviderService",
+
+ # /System/Library/Frameworks
+ "/System/Library/Frameworks/CoreML.framework/CoreML",
+ "/System/Library/Frameworks/CoreVideo.framework/CoreVideo",
+ "/System/Library/Frameworks/ManagedAppDistribution.framework/Support/managedappdistributiond",
+ "/System/Library/Frameworks/MediaToolbox.framework/MediaToolbox",
+ "/System/Library/Frameworks/SoundAnalysis.framework/SoundAnalysis",
+ "/System/Library/Frameworks/StoreKit.framework/Support/storekitd",
+
+ # /System/Library/PrivateFrameworks
+ "/System/Library/PrivateFrameworks/AAAFoundation.framework/AAAFoundation",
+ "/System/Library/PrivateFrameworks/AirPlaySupport.framework/AirPlaySupport",
+ "/System/Library/PrivateFrameworks/AppStoreDaemon.framework/Support/appstored",
+ "/System/Library/PrivateFrameworks/AppStoreUtilities.framework/AppStoreUtilities",
+ "/System/Library/PrivateFrameworks/AppleNeuralEngine.framework/AppleNeuralEngine",
+ "/System/Library/PrivateFrameworks/ApplePushService.framework/ApplePushService",
+ "/System/Library/PrivateFrameworks/ApplePushService.framework/apsd",
+ "/System/Library/PrivateFrameworks/AuthKit.framework/AuthKit",
+ "/System/Library/PrivateFrameworks/CMCapture.framework/CMCapture",
+ "/System/Library/PrivateFrameworks/CoreALD.framework/CoreALD",
+ "/System/Library/PrivateFrameworks/CoreCDP.framework/CoreCDP",
+ "/System/Library/PrivateFrameworks/CorePrescription.framework/CorePrescription",
+ "/System/Library/PrivateFrameworks/CorePrescription.framework/XPCServices/CorePrescriptionService.xpc/CorePrescriptionService",
+ "/System/Library/PrivateFrameworks/CoreRE.framework/CoreRE",
+ "/System/Library/PrivateFrameworks/DVTInstrumentsUtilities.framework/DVTInstrumentsUtilities",
+ "/System/Library/PrivateFrameworks/DataDetectorsUI.framework/PlugIns/com.apple.DataDetectorsUI.ActionsExtension.appex/com.apple.DataDetectorsUI.ActionsExtension",
+ "/System/Library/PrivateFrameworks/DeviceCheckInternal.framework/DeviceCheckInternal",
+ "/System/Library/PrivateFrameworks/DeviceIdentity.framework/DeviceIdentity",
+ "/System/Library/PrivateFrameworks/EmailFoundation.framework/EmailFoundation",
+ "/System/Library/PrivateFrameworks/Espresso.framework/Espresso",
+ "/System/Library/PrivateFrameworks/FindMyBase.framework/FindMyBase",
+ "/System/Library/PrivateFrameworks/IDS.framework/identityservicesd.app/identityservicesd",
+ "/System/Library/PrivateFrameworks/IDSFoundation.framework/IDSFoundation",
+ "/System/Library/PrivateFrameworks/IOSurfaceAccelerator.framework/IOSurfaceAccelerator",
+ "/System/Library/PrivateFrameworks/MagnifierSupport.framework/MagnifierSupport",
+ "/System/Library/PrivateFrameworks/MobileActivation.framework/MobileActivation",
+ "/System/Library/PrivateFrameworks/NeuralNetworks.framework/NeuralNetworks",
+ "/System/Library/PrivateFrameworks/PhotoFoundation.framework/PhotoFoundation",
+ "/System/Library/PrivateFrameworks/Recon3D.framework/Recon3D",
+ "/System/Library/PrivateFrameworks/RenderBox.framework/RenderBox",
+ "/System/Library/PrivateFrameworks/SonicKit.framework/SonicKit",
+ "/System/Library/PrivateFrameworks/TrialServer.framework/TrialServer",
+ "/System/Library/PrivateFrameworks/VFX.framework/VFX",
+ "/System/Library/PrivateFrameworks/VideosUI.framework/VideosUI",
+ "/System/Library/PrivateFrameworks/VisionKitCore.framework/VisionKitCore",
+ "/System/Library/PrivateFrameworks/WatchdogServiceManagement.framework/WatchdogServiceManagement",
+ "/System/Library/PrivateFrameworks/caulk.framework/caulk",
+
+ # /Applications
+ "/Applications/CheckerBoard.app/CheckerBoard",
+ "/Applications/Family.app/Family",
+ "/Applications/PeopleMessageService.app/PeopleMessageService",
+ "/Applications/PeopleViewService.app/PeopleViewService",
+ "/Applications/StoreKitUISceneService.app/StoreKitUISceneService",
+)
+
+
+# ─────────────────────────────────────────────────────────────────────
+# Rootfs binaries to LEAVE UNPATCHED.
+#
+# A path in this set keeps its `kern.hv_vmm_present\0` cstring as-is.
+# After the kernel rename that sysctl name returns ENOENT, the caller's
+# defensive post-call check leaves the cached `is_vmm` byte at BSS-zero
+# (0), and the binary thinks "not in a VM". Use this for sign-in /
+# Apple-ID / activation / Store consumers — mirrors the DSC blacklist.
+#
+# Starting set is intentionally conservative: only daemons we have a
+# clear "is sign-in adjacent" reason to leave alone. Tighten / loosen
+# by editing this list and re-running `make cfw_install_jb`.
+# ─────────────────────────────────────────────────────────────────────
+DONT_PATCH_ROOTFS_PATHS: frozenset[str] = frozenset((
+ # Apple ID / activation / device-identity daemons
+ "/usr/libexec/adid",
+ "/usr/libexec/mobileactivationd",
+ "/usr/libexec/fairplaydeviceidentityd",
+ "/System/Library/DataClassMigrators/MobileActivationMigrator.migrator/MobileActivationMigrator",
+ # APNS / iMessage daemons
+ "/System/Library/PrivateFrameworks/ApplePushService.framework/apsd",
+ "/System/Library/PrivateFrameworks/IDS.framework/identityservicesd.app/identityservicesd",
+ # Store / IAP daemons + UI
+ "/System/Library/Frameworks/StoreKit.framework/Support/storekitd",
+ "/System/Library/PrivateFrameworks/AppStoreDaemon.framework/Support/appstored",
+ "/Applications/StoreKitUISceneService.app/StoreKitUISceneService",
+ "/Applications/CheckerBoard.app/CheckerBoard",
+ # Health Rx
+ "/System/Library/PrivateFrameworks/CorePrescription.framework/XPCServices/CorePrescriptionService.xpc/CorePrescriptionService",
+ # Standalone copies of DSC-blacklisted frameworks. The rootfs files
+ # may be stubs that redirect into the DSC at load time, in which
+ # case mangling them has no effect — but blacklisting is defensive
+ # and matches role intent.
+ "/System/Library/PrivateFrameworks/AAAFoundation.framework/AAAFoundation",
+ "/System/Library/PrivateFrameworks/AuthKit.framework/AuthKit",
+ "/System/Library/PrivateFrameworks/IDSFoundation.framework/IDSFoundation",
+ "/System/Library/PrivateFrameworks/DeviceIdentity.framework/DeviceIdentity",
+ "/System/Library/PrivateFrameworks/DeviceCheckInternal.framework/DeviceCheckInternal",
+ "/System/Library/PrivateFrameworks/MobileActivation.framework/MobileActivation",
+ "/System/Library/PrivateFrameworks/ApplePushService.framework/ApplePushService",
+ "/System/Library/PrivateFrameworks/AppStoreUtilities.framework/AppStoreUtilities",
+ "/System/Library/PrivateFrameworks/CorePrescription.framework/CorePrescription",
+ "/System/Library/PrivateFrameworks/CoreCDP.framework/CoreCDP",
+ "/System/Library/PrivateFrameworks/EmailFoundation.framework/EmailFoundation",
+ "/System/Library/PrivateFrameworks/FindMyBase.framework/FindMyBase",
+ "/System/Library/PrivateFrameworks/TrialServer.framework/TrialServer",
+ "/System/Library/PrivateFrameworks/DVTInstrumentsUtilities.framework/DVTInstrumentsUtilities",
+ "/System/Library/PrivateFrameworks/WatchdogServiceManagement.framework/WatchdogServiceManagement",
+))
+
+
+def get_patch_paths() -> list[str]:
+ """Return the list of rootfs binary paths that should be byte-5
+ mangled at install time. This is `ALL_KNOWN_ROOTFS_PATHS` minus
+ `DONT_PATCH_ROOTFS_PATHS`, in stable sorted order.
+ """
+ return sorted(p for p in ALL_KNOWN_ROOTFS_PATHS if p not in DONT_PATCH_ROOTFS_PATHS)
+
+
+def main(argv: list[str]) -> int:
+ """Print the patch paths, one per line, for shell consumption."""
+ for p in get_patch_paths():
+ print(p)
+ return 0
+
+
+if __name__ == "__main__":
+ raise SystemExit(main(sys.argv))