Bump version to 0.1.6

It hard-coded VR_X, so a target whose page tables expose no executable
region silently yielded zero xrefs. Callers now pass the scope (VR_X to
restrict to code, VR_R when the X bit is unavailable); the decoder and
exact target match remain the correctness gate. Mirrors the prot_any
parameter gva_imm_xref already carries.

.gitea/workflows/release.yml

@@ -86,7 +86,7 @@ jobs:
         run: |
           apt-get update
           apt-get install -y --no-install-recommends \
-            cmake make gcc libc6-dev dpkg-dev \
+            cmake make gcc libc6-dev dpkg-dev file \
             gcc-mingw-w64-x86-64 ca-certificates curl
 
       - name: Configure

CMakeLists.txt

@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.18)   # find_program(... REQUIRED)
 
-set(VMIE_VERSION "0.1.4" CACHE STRING "Library version (MAJOR.MINOR.PATCH); CI passes the tag")
+set(VMIE_VERSION "0.1.6" CACHE STRING "Library version (MAJOR.MINOR.PATCH); CI passes the tag")
 project(vmi-engine VERSION ${VMIE_VERSION} LANGUAGES C)
 
 set(CMAKE_C_STANDARD 17)               # generation B uses no C23 feature

include/memmodel.h

@@ -125,7 +125,11 @@ typedef struct {
  * x86-64 has no read bit: a present page is readable, so VR_R is always set on a
  * returned region. Write/execute/user are the EFFECTIVE rights along the whole
  * page-table path (RW & US are AND-ed across levels, NX is OR-ed), not just the
- * leaf entry, so they reflect what the guest CPU actually enforces. */
+ * leaf entry, so they reflect what the guest CPU actually enforces. One
+ * exception keeps VR_X honest under KVA shadow (KPTI): the kernel CR3 marks
+ * user-half PML4Es NX as hardening, but ring 3 executes those pages via the
+ * user/shadow CR3 (same shared lower tables, NX-clear PML4E), so the user-half
+ * top-level NX bit is ignored when deriving VR_X. */
 #ifndef VMIE_VREGION_DEFINED
 #define VMIE_VREGION_DEFINED
 #define VR_R 0x1u   /* readable  (present => always set)            */

include/scan.h

@@ -70,18 +70,21 @@ int gva_sig_scan_multi(vmie_mem* m, uintptr_t cr3, uint64_t lo, uint64_t hi,
                        uint32_t prot_any, const sigset* s,
                        sig_multi_hit* out, int max);
 
-/* code-xref: every instruction in the X-regions of [lo,hi] whose near rel
+/* code-xref: every instruction in the regions of [lo,hi] kept by `prot_any`
- * branch or RIP-relative memory operand resolves to `target_va`. Brute-scans
+ * whose near rel branch or RIP-relative memory operand resolves to `target_va`.
- * each byte offset with the light x86-64 decoder (x86dec.h, NOT a full
+ * Brute-scans each byte offset with the light x86-64 decoder (x86dec.h, NOT a
- * disassembler): an E8/E9/EB/Jcc rel branch matches when next_rip + rel ==
+ * full disassembler): an E8/E9/EB/Jcc rel branch matches when next_rip + rel ==
  * target_va, and any RIP-relative operand (ModRM mod=00, rm=101) matches when
  * next_rip + disp32 == target_va (this covers lea/mov and any other rip-rel
- * form). Records each matching instruction-start VA. The sweep forces VR_X and
+ * form). Records each matching instruction-start VA. The sweep is scoped by
- * carries a >=15-byte overlap (max x86 instruction length) so no instruction is
+ * `prot_any` (pass VR_X to restrict to code; pass VR_R when the target's
- * cut at a window seam. Writes up to `max` VAs to `out` (NULL to count only) and
+ * page-table X bit is unavailable - the decoder and exact target match keep the
- * returns the TOTAL number of matches, or -1 on bad input. */
+ * result correct, only more bytes are decoded) and carries a >=15-byte overlap
+ * (max x86 instruction length) so no instruction is cut at a window seam. Writes
+ * up to `max` VAs to `out` (NULL to count only) and returns the TOTAL number of
+ * matches, or -1 on bad input. */
 int gva_code_xref(vmie_mem* m, uintptr_t cr3, uint64_t lo, uint64_t hi,
-                  uint64_t target_va, uint64_t* out, int max);
+                  uint32_t prot_any, uint64_t target_va, uint64_t* out, int max);
 
 /* immediate / constant xref: every instruction in [lo,hi] (kept by the
  * protection filter `prot_any`; pass VR_X to restrict to code) whose IMMEDIATE

src/engine/gva.c

@@ -158,7 +158,16 @@ int gva_regions(vmie_mem* m, uintptr_t cr3, uint64_t lo, uint64_t hi,
         if (!(e4 & PG_P)) continue;
         const uint64_t b4 = VA_CANON((uint64_t)i4 << 39);
         if (!rgn_hit(b4, 1ull << 39, lo, hi)) continue;
-        const int rw4 = (e4 >> 1) & 1, us4 = (e4 >> 2) & 1, nx4 = (int)(e4 >> 63) & 1;
+        const int rw4 = (e4 >> 1) & 1, us4 = (e4 >> 2) & 1;
+        /* On a KVA-shadow (KPTI) Windows the kernel CR3 marks user-half PML4Es
+         * NX as hardening: the kernel must not execute user pages through its own
+         * mapping. Ring 3 runs under the user/shadow CR3, which reaches the SAME
+         * lower tables (PDPT/PD/PT are shared) through a PML4E with NX clear, so
+         * that top-level NX is not what executes user code. Drop it on the user
+         * half; sub-PML4E NX stays authoritative (it is shared between the two
+         * CR3s). On a non-KPTI guest a code-covering user PML4E already has NX
+         * clear, so this only ever discards a zero. i4 < 256 == canonical user. */
+        const int nx4 = (i4 < 256) ? 0 : ((int)(e4 >> 63) & 1);
 
         const uint64_t* t3 = gpa_ptr(m, e4 & PFN_MASK, 4096);
         if (!t3) continue;

src/handlers/codescan.c

@@ -115,10 +115,10 @@ static int xref_sweep_cb(void* u, const uint8_t* data, size_t len,
 }
 
 int gva_code_xref(vmie_mem* m, uintptr_t cr3, uint64_t lo, uint64_t hi,
-                  uint64_t target_va, uint64_t* out, int max) {
+                  uint32_t prot_any, uint64_t target_va, uint64_t* out, int max) {
     struct xref_cb c; memset(&c, 0, sizeof c);
     c.target = target_va; c.out = out; c.max = max;
-    if (gva_sweep(m, cr3, lo, hi, VR_X, X86_MAX_INSN, xref_sweep_cb, &c) < 0) {
+    if (gva_sweep(m, cr3, lo, hi, prot_any, X86_MAX_INSN, xref_sweep_cb, &c) < 0) {
         return -1;
     }
     return c.n;