fix(relinker): refuse branches into trapped guest GS instructions on Windows

Windows output replaced each guest GS memory instruction with UD2 and
NOP fill without checking the decoded direct branch targets, so a branch
past the first byte of such an instruction was converted and landed in
the second UD2 byte or the NOP fill. A branch target strictly inside a
trapped GS instruction is now refused with "Branch enters a guest GS
instruction", the rule macOS applies since #7706 and this builder
already applies to its FS rewrites. The branch targets are sorted before
the trap loop, which runs before the PT_TLS early returns, so images
without PT_TLS are covered; guest modules go through the same builder. A
branch to the first byte still converts and reaches the UD2.

core/relinker/elfpatcher/src/windows/WindowsTlsBuilder.cpp

@@ -221,8 +221,14 @@ PeDirectory WindowsTlsBuilder::Build(const std::vector<std::uint8_t>& source, co
     }
 
     report(9000);
-    for (const auto& [rva, length] : gsTraps)
+    std::sort(branchTargets.begin(), branchTargets.end());
+    branchTargets.erase(std::unique(branchTargets.begin(), branchTargets.end()), branchTargets.end());
+    for (const auto& [rva, length] : gsTraps) {
+        const auto target = std::upper_bound(branchTargets.begin(), branchTargets.end(), rva);
+        if (target != branchTargets.end() && *target < rva + length)
+            throw Domain::RelinkerException("Branch enters a guest GS instruction", *target);
         trapInstruction(sections, rva, length);
+    }
     if (tls == nullptr) {
         if (!accesses.empty())
             throw Domain::RelinkerException("Guest TLS access without PT_TLS");
@@ -274,9 +280,6 @@ PeDirectory WindowsTlsBuilder::Build(const std::vector<std::uint8_t>& source, co
     code.PatchBranch(skipCallback, code.GetRva());
     code.Emit({0xc3});
 
-    std::sort(branchTargets.begin(), branchTargets.end());
-    branchTargets.erase(std::unique(branchTargets.begin(), branchTargets.end()), branchTargets.end());
-
     std::uint64_t patched = 0;
     for (const auto& access : accesses) {
         if (patched % 256 == 0) report(9000 + 900 * patched / accesses.size());

core/relinker/relinker/tests/test_tls_function_coverage.py

@@ -777,6 +777,8 @@ def main():
             conflicting[232:344] = unwind[232:344]
             struct.pack_into("<H", conflicting, 56, 5)
             convert("conflicting-function-extents", conflicting, "Code analysis: conflicting function ranges")
+            convert("branch-into-gs", make_image("register", "unwind", body=bytes.fromhex("74 01 65 48 8b 0c 25 78 03 00 00 c3")),
+                    "Branch enters a guest GS instruction", error_offset=0x11243)
 
         def check_gs_traps(cases):
             for name, instruction, trapped in cases: